presentation at the joint unece/eurostat/unfpa/medstat ii work session on migration statistics,...

Presentation at the Presentation at the Joint UNECE/Eurostat/UNFPA/MEDSTAT II Joint UNECE/Eurostat/UNFPA/MEDSTAT II

Work Session on Migration Statistics, Work Session on Migration Statistics, Geneva, 3-5 March 2008Geneva, 3-5 March 2008

Michael Jandl

“Methodologies for the estimation of stocks of irregular migrants"

Michael Jandl, Geneva, 4.3.2008 2

Outline of presentation

1. Research and data problem2. A classification of methods for estimating irregular migrant

stocks and flows3. Selected methods for estimating irregular migrant stocks4. Conclusions


Research and data problem

Illegal (unauthorized/undocumented) Migration

- is deliberately hidden from view (sanctioned)- is statistically not directly quantifyable

- can be researched with qualitative social science methods, data trend analysis and quantitative estimation methods

- Data and methods should be subject to close scrutiny

- Minimum condition: indications on methods, assumptions and data sources of quantitative estimates should be given

- Otherwise we should not speak of an „estimate“


A new classification scheme

Differentiation:- Stock data (illegal residence, illegal work)- Flow data (illegal entry)

Subdivisions into:- Direct approaches vs. indirect approaches (+ combined approaches)

- Data sources (already existing, specifically generated)- Methodic grouping

- Estimation technique


A classification of methods for estimating irregular migrant stocks I

Approach Data sources Method Estimation technique

Direct approaches

Based on immigration enforcement data

Multiplier methods Simple Multiplier

Capture-recapture

Repeated capture

Matching of registers

Statistical methods Random effect mixed modelling approach

Based on administrative statistics

Methods of self-identification

Evidence based on regularisation data

Using data on status adjustments over time

Based on surveys Snowball sampling Direct survey methods

Single stage link-trace sampling


A classification of methods for estimating irregular migrant stocks II


Indirect approaches

Based on census/registers

Residual methods Differences census results – legal immigration data

Simple comparison of various registers

Demographic methods Use of birth/death rates

Based on surveys of “key informants”

Subjective Estimations/ Indicators Methods

Expert surveys

Delphi surveys

Based on non-demographic data

Econometric methods on shadow economy

Inference from estimates on illegal work

Based on census/registers/ demographic data

Expected population methods

Comparison of census/emigration data and immigration statistics

Based on admin. statistics

Flow-stock methods Calculating the stock through flow figures

Based on complem. data sources

Indirect inferences Registered school children, household surveys, etc.


A classification of methods for estimating irregular migrant stocks III


Combined approaches

Based on small scale surveys

Window/Postal code method

Small scale study / use of regression analysis

Based on expert opinions

Localized Delphi Delphi method / use of regression analysis

Adjustment to surveys/ census data

Non-threatening survey design

Randomized response (3 cards method) / residual method


A classification of methods for estimating irregular migrant flows

Approach Data sources Method Model

Direct approaches

Based on border apprehension data

Multiplier methods Simple multiplier

Indirect approaches

Based on stock estimates

Differential methods Net differences in stocks

Based on entry-exit statistics

Residual method Double entry card system


Multiplier Method– Based on the projection of available indicators, using an appropriately

defined/derived multiplier.. on stocks/flows

- Flow data (illegal entries): e.g. On apprehension data of illegal migrants at the border

- Stock data (illegal residence, illegal work)

Example: Burgers 1995Based on the number of apprehended criminal foreignersIn-depth interviews with a sample of illegal migrants in Rotterdam to determine share of

illegal migrants involved in criminal activities multiplier estimate of total

Methods for estimating irregular migrant stocks I


Repeated Capture method in the NLVan der Leun et al. (1998), Engbersen et al. (2002), Van der Heijden (2006)

- Analysis of data from police enforcement records from 25 police districts- Establishing who has been caught 1, 2, 3, … times- Fitting the data to a Poisson distribution and calculate N (y=0)

Estimated Illegal Immigrants never apprehended (y=0) in repeated capture model

0

2000

4000

6000

8000

10000

12000

Reihe1 10810 1645 183 37 13 1 1

0 1 2 3 4 5 6

Methods for estimating irregular migrant stocks II


Methods for estimating irregular migrant stocks III

A random effect mixed modelling approach

This is currently tested for Norway (UDI project, Li-Chun Zhang (2007)

N= known/registered populationn= apprehended by the police but among NM= unknown/unregistered population to be estimatedm= those apprehended among MN(i), n(i), M(i), m(i) are subgroups by age, citizenship,...

1) M(i)/N(i) is a random deviation from a global proportionality coefficient: log(Mi/Ni) = + vi

2) (mi)/M(i) is a function of n(i)/N(i): log(mi/Mi) = log(ni/Ni) + ei

Mi can be estimated by OLS and M= Mi for all i


Evidence from regularizations (examples)Italy’s 2002/2003 regularization:- 700,000 applications - 640,000 regularizations- Top 3 countries were: Romania, Ukraine, AlbaniaSpain’s 2005 regularization:- 690,000 applications – 573,000 positive (end 05)- Top 3 countries were: Ecuador, Romania, Morocco2004/2007 EU enlargement = regularization?:

Problems: - Number of applications is not the same as number of persons- Wide differences in implementation (Italy – easier; Greece – difficult)- Not all illegal residents apply; additional non-residents apply- Persons regularized can fall back into irregularity

Methods for estimating irregular migrant stocks IV


Methods for estimating irregular migrant stocks V

Single-stage link-tracing sampling

Frank and Snijders (1994), Li-Chun Zhang (2007)- A specific form of snowball sampling,..- Taking a random sample s (1,...,i,...n) of persons in a target population U

(1,...N) and letting each person (i) nominate m(i) other persons among U, then r(i) denotes the number of persons referring back to sample s.

Based on this, under certain assumptions an estimator can be obtained, so that N (est)= n + (n-1)x(m-r)/r

(that is the higher the number of persons already nominated by i, the smaller the total population N)


„Residual methods“

Based on the differences between - Census results (total population adjusted for undercounting) and- Immigration data, aliens registers, residence permit registers, etc.

- Example USA: 2005 estimate: 10.3 m (Passel 2005), Jan 2006 estimate: 11.6m (Hoefer et al); UK: 2001 estimate: 310k-430k-570k

Methods for estimating irregular migrant stocks VI


„Simple Comparison of Registers“

Indirect estimation based on a comparison of two or more registers with data on the same target population (where irregular migrants may at least partially be included)

Example Spain:Legal residents should be registered in database on residents permits while

illegal residents have incentives to register in the municipal register

Methods for estimating irregular migrant stocks VII

Register 1998 1999 2000 2001 2002 2003 2004 2005

Municipal Register 637.085 748.954 923.879 1.370.657 1.977.946 2.664.168 3.034.326 3.730.610

Foreigners with Resident Permit

719.647 801.332 895.720 1.109.060 1.324.001 1.647.011 1.977.291 2.738.932

Difference (MR – Foreigners with Resident Permit

- 82.562 -52.378 28.159 261.597 653.945 1.017.157 1.057.035 991.678


„Comparison of Immigration and Emigration Statistics“

Based on the differences between (estimated) emigration and (recorded) immigration data, the number of migrants without legal resident status is estimated

Examples: Mexico-US migration (in: Lederer 1994); Delaunay and Tapinos (1998a): Morocco – Europe (not Tunisia), Fargues (2007): estimates of MENA emigration; Malynovska 2004: estimates on Ukrainians abroad

Problems: - Migrants may disperse over a large number of countries not just to one region- Both emigration estimates and immigration data are prone to large margins of

error- However, data from countries of origin/consulates could be more exploited

Methods for estimating irregular migrant stocks VIII


Conclusions I

1. There are a number of feasible estimation methods available2. A lot can be learned from existing examples in migration research

and from the study of other hidden populations3. Not all methods are suitable and applicable in all circumstances

and at all times4. Some methods need the production of new data, some can use

existing data (if available)5. In addition to data and methods, a good knowledge of irregular

migration processes is necessary ( also qualitiative research!)6. A solid estimate usually needs substantial time and resources


Conclusions II

1. No one estimate can be perfect, each method has its strengths and weaknesses

2. An external validation of the accuracy of results is difficult3. A triangulation of several independent estimates can improve

confidence in the outcome4. The use of multiple, complementary methods is recommendable5. New methods and data sources can and should be developed6. Technological and administrative innovations can give rise to new

methods (e.g. Biometric visa and travel documents for entry-exit method or electronic databases with fingerprints for repeated capture method)

presentation at the joint unece/eurostat/unfpa/medstat ii work session on migration statistics,...

Documents