29.10.2014 Views

Educational Research - the Ethics and Aesthetics of Statistics

Educational Research - the Ethics and Aesthetics of Statistics

Educational Research - the Ethics and Aesthetics of Statistics

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

100 E. Hemelsoet<br />

a very informative schematic overview, see Appendixes 1 <strong>and</strong> 2). Three examples<br />

are given though, as <strong>the</strong>y are paradigmatic for <strong>the</strong> problems related to estimating<br />

irregular migrants.<br />

Multiplier methods calculate <strong>the</strong> size <strong>of</strong> <strong>the</strong> unknown total from <strong>the</strong> size <strong>of</strong> a<br />

known subtotal by use <strong>of</strong> an appropriately estimated multiplier. The use <strong>of</strong> multipliers<br />

to derive <strong>the</strong> size <strong>of</strong> a hidden population from <strong>the</strong> size <strong>of</strong> a known subtotal<br />

<strong>of</strong> that population is probably <strong>the</strong> most common way <strong>of</strong> estimating an unknown<br />

population. The problem <strong>of</strong> giving adequate estimations is <strong>the</strong>n translated in finding<br />

<strong>the</strong> right multiplier (Vogel, 2002). An example <strong>of</strong> using <strong>the</strong> multiplier method<br />

for estimating irregular migrant numbers is provided by a recent study in Belgium<br />

by Van Meeteren, Van San, <strong>and</strong> Engbergsen (2007). In this study, police data on<br />

arrested criminal foreigners is combined with data from in-depth interviews with<br />

120 irregular migrants. The police data on criminal <strong>of</strong>fences committed by foreigners<br />

without legal residence is compared with <strong>the</strong> ‘crime rate’ among <strong>the</strong> migrants<br />

that were interviewed. Three assumptions are internal to this method, assumptions<br />

that modify <strong>the</strong> quality <strong>of</strong> <strong>the</strong> outcomes. First, <strong>the</strong>re is <strong>the</strong> expectation that <strong>the</strong> crime<br />

rate derived from <strong>the</strong> interview sample is a good indicator for <strong>the</strong> actual crime rate<br />

among <strong>the</strong> target population. Second, <strong>the</strong> estimate is based on <strong>the</strong> assumption that<br />

<strong>the</strong> reported crime figures are a good indicator for total crime figures. Thirdly, <strong>the</strong><br />

sampling group is expected to be representative <strong>of</strong> <strong>the</strong> total population <strong>of</strong> irregular<br />

migrants. Concerning <strong>the</strong> first assumption, it can be argued that long in-depth<br />

interviews lead to sufficient trust between <strong>the</strong> interviewer <strong>and</strong> <strong>the</strong> respondent so as<br />

to retrieve realistic answers. As regards <strong>the</strong> second assumption, things get slightly<br />

more complicated. It is not clear to what extent police data on arrests cover <strong>the</strong><br />

total amount <strong>of</strong> criminal <strong>of</strong>fences that are actually committed. It is very likely that<br />

<strong>the</strong> police do not detect all cases <strong>of</strong> crime <strong>and</strong> <strong>the</strong> resulting estimate is <strong>the</strong>refore<br />

unreliable (Kraler & Vogel, 2008). The third assumption implies problems concerning<br />

generalisation, as <strong>the</strong> sampling group is ra<strong>the</strong>r small <strong>and</strong> it is impossible<br />

to retrieve a r<strong>and</strong>omised sample if <strong>the</strong> composition <strong>of</strong> <strong>the</strong> total group is largely<br />

unknown.<br />

Similar problems arise with <strong>the</strong> capture-recapture method. This method is used<br />

in biology to estimate animal populations in <strong>the</strong> wild. The basic idea is to develop a<br />

multiplier through repeated sampling <strong>of</strong> <strong>the</strong> same population. Estimating a fish population<br />

goes as follows: first, capture 1,000 fish <strong>and</strong> mark <strong>the</strong>m before releasing <strong>the</strong>m<br />

again. Capture ano<strong>the</strong>r 1,000 fish <strong>and</strong> look how many are marked. If for example 100<br />

are marked, <strong>the</strong> 1,000 fish will statistically make up approximately 10% <strong>of</strong> <strong>the</strong> total,<br />

i.e. 10,000 fish. In <strong>the</strong> Ne<strong>the</strong>rl<strong>and</strong>s Van der Leun, Engbersen, <strong>and</strong> van der Heijden<br />

(1998) have used this principle to estimate numbers <strong>of</strong> irregular migrants. Their<br />

‘repeated capture method’ is based on a dataset that tries to apprehend numbers <strong>of</strong><br />

illegal immigrants in Amsterdam, Rotterdam, The Hague <strong>and</strong> Utrecht, (<strong>the</strong> four big<br />

Dutch cities) during 1995. Using <strong>the</strong> number <strong>of</strong> persons captured <strong>and</strong> arrested again,<br />

it is argued that <strong>the</strong> number <strong>of</strong> people who will show up again follows a probabilistic<br />

distribution called <strong>the</strong> Poisson distribution (λ). On <strong>the</strong> basis <strong>of</strong> <strong>the</strong> available data,<br />

<strong>the</strong> crucial parameter determining <strong>the</strong> Poisson distribution can be estimated. This

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!