You are on page 1of 33

27

Cross-Nationally Comparative Research


on Racial and Ethnic Skill Disparities:
Questions, Findings, and Pitfalls
Alexander Dicks, Jaap Dronkers†, and Mark Levels

Introduction
This review describes and assesses how educational and social scientists have
used large-scale cross-national assessment data to study racial and ethnic
inequalities in primary and secondary education between 2000 and 2017.
Previous reviews of immigrant children’s educational performance focused
on a distinct origin group (Crul and Vermeulen 2003) or on a single coun-
try (Kao and Thompson 2003). Although several review studies pay atten-
tion to findings from assessment data for studying immigrant children’s
educational performance, none were written with the distinct purpose of
discussing how the availability of these large scale data sets has contributed
to the study of immigrant children’s performance (i.e. Alba et al. 2011;
Heath et al. 2008). To fill this gap, we provide an overview of studies that

A. Dicks
Department of Economics, Maastricht University, Maastricht, The Netherlands
e-mail: a.dicks@maastrichtuniversity.nl
J. Dronkers†
Deceased
M. Levels (*)
Research Centre for Education and the Labor Market (ROA), Maastricht
University, Maastricht, The Netherlands
e-mail: m.levels@maastrichtuniversity.nl

© The Author(s) 2019 1183


P. A. J. Stevens, A. G. Dworkin (eds.), The Palgrave Handbook of Race and Ethnic Inequalities
in Education, https://doi.org/10.1007/978-3-319-94724-2_27
1184 A. Dicks et al.

used large-scale assessment data (i.e. PISA, PIRLS, and TIMMS). We also
add a distinct focus on origin country differences, whereas for example Alba
et al. (2011) focused more on differences between destination countries. We
identify the type of research questions that can be addressed with such data,
discuss the main empirical findings, and identify the main short-comings of
and conclusions from this literature.
We argue that the availability of consecutive waves of large-scale assessment
data has led to an evolution of research questions that can be answered with
cross-national assessment data. These questions are:

1. How does the relation between individual-level background characteristics


and performance of immigrant and native pupils on assessment tests vary
between destination countries?
2. How does the relation between school-level characteristics and perfor-
mance of immigrant and native pupils on assessment tests vary between
destination countries?
3. What is the relation between characteristics of destination countries and
immigrant children’s performance on assessment tests?
4. What is the relation between origin country traits and immigrant chil-
dren’s performance on assessment tests?

These questions have fed a flourishing fundamental research agenda, which


we will describe in this chapter. However, it should be noted that answering
these questions with data on immigrant children’s scores on internationally
comparable achievement tests has important social and political implications.
There has been a strong tendency amongst policy-makers to view high-quality
education as a necessary condition for economic growth (see: Pigozzi 2006).
Test scores from cross-national achievement surveys have been used and are
being used to assess how countries’ education systems succeed in teaching
certain skills deemed elementary for economic and societal participation,
such as math and reading. PISA is actually explicitly designed for this purpose
(Schleicher 1999). Country assessments are usually made by ranking coun-
tries based on their pupils’ average score on the tests (see: OECD 2015a, for
a recent example). Such rankings are taken very seriously by policy-makers
and politicians. Famously, the publication of the first PISA ranking in 2000
lead to a veritable PISA-shock in Germany, whose politicians deemed the rela-
tively low ranking of Germany unacceptable (Gruber 2006, as well as the
chapter on Germany in this Handbook). PISA has sparked similar debates in
other countries (Breakspear 2012, see also chapters on Italy and Israel in this
Handbook). Relatively low PISA scores could affect the willingness of
Cross-Nationally Comparative Research on Racial and Ethnic Skill… 1185

­ oliticians to accept immigrants from certain origin countries. For example,


p
German politician Thilo Sarrazin cited a table from Levels et al.’s (2008) anal-
yses of PISA data to support an argument against immigration of low-skilled
immigrants from predominantly Muslim countries to Germany (Sarrazin
2010: p. 368). By improving our understanding of what drives immigrant-­
native skills gaps, researchers can provide these public debates with the neces-
sary qualifications and nuance.
The rest of this chapter is divided in five parts. First, in section “Analyses of
Immigrant Children’s Performance on International Assessment Data”, we
shortly provide arguments for using large-scale assessment data to study the
educational performance of immigrant children. We show how large-scale
assessment data can be used to study questions related to ethnicity and race,
education systems, and social policies. We argue how studying such data helps
us to advance on single country studies, and discuss some of the potential
pitfalls. In the third section, we explain the process of this review. We explain
the search strategies we employed and give detailed information about inclu-
sion criteria for this review. The fourth section forms the core of this chapter.
Here, we turn from potential contributions to discussing how these data were
actually used. To do this, we provide an overview of relevant studies. In the
fifth section, we discuss a number of issues and problems with cross-national
analyses of immigrant children’s performance, and provide some critical sug-
gestions about the way in which literature should develop. In the final section,
we draw some general conclusions.

 nalyses of Immigrant Children’s Performance


A
on International Assessment Data
Immigrants, Race and Ethnicity in Cross-National
Education Surveys

Questions on the role of race and ethnicity have long been at the core of
research on immigrant children’s educational performance. Traditionally, such
research was done in single countries (mostly the US), where assessment data
were used to examine differences in achievement between children from dif-
ferent ethnic backgrounds. As we will show in this chapter, the use of large-­
scale assessment data to address such questions has provided some remarkable
insights about the generalizability and explanations of some observations of
the performance of origin groups in single countries. For example, it has long
been established that immigrant children from South-East Asia perform
1186 A. Dicks et al.

remarkably well in US education. Cross-national research using PISA has


shown that this observation can be generalized to other countries as well
(Levels and Dronkers 2008). Similarly, while it was long inferred from coun-
try case studies that there is cross-national variation in the extent to which
young Turks perform in school (Crul and Vermeulen 2003), an attempt to
test explanations for such destination country differences in origin group
performance must rely on comparable cross-national data (cf. Levels et al.
2008).
The way in which cross-national assessment data can provide variation to
explain, is illustrated by Table 27.1. This table shows how – on average –
immigrant children of certain origin countries (rows) in certain destination
countries (columns) score on the PISA 2003 reading literacy test. In many
countries, immigrants (on average) perform worse than natives. The table also
shows that there is extensive variation between and within destination coun-
tries, and between and within origin countries. For example, we observe that
immigrant children in Australia and New Zealand perform on par with
or better than native children. In Austria and Germany, however, immigrant
children are performing less well than natives. Looking within destination
countries, we see that Albanians in Switzerland perform worse than Serbians
in Switzerland. Between origin countries, however, we see that Serbians in
Switzerland are outperformed by Serbians in Germany. Turks in Germany,
however, are largely comparable to Turks in Austria, yet both achieve lower
test scores than Turks in the Netherlands. Other comparisons are informative
as well. For instance, Germans in Australia outperform Germans in Germany,
whereas New Zealanders in Australia underperform compared to New
Zealanders in New Zealand. Large-scale assessment data helps to tease out
this variation, i.e. by comparing the same origin groups in different destina-
tion countries and at the same time comparing different origin groups in the
same destination country. Because the measurement of reading literacy is
identical in the countries, the observed cross-national differences can likely be
attributed to actual achievement differences. Apparently there are differences
between the countries that make immigrants achieve at different levels.
Explaining these differences is an important goal of this entire research line.
Using cross-national data to study origin differences delivers valuable
insights, but it is not without downsides. One major issue is that origin coun-
try data is not available for some countries. This has two reasons. First, many
countries have legal provisions that outlaw asking about race, ethnicity, or
origin country in surveys. Some countries refrain from addressing these issues
altogether, other countries use a broad coding scheme that for example only
distinguishes between western and non-western immigrants. OECD member
Table 27.1 Average reading literacy of immigrant pupils per country of destination and country of origin (N = 7.459)
Countries of destination
Countries of origin AU AT BE CH DE DK EL IE LV LU NL NZ SC Total
Albania 368 377 424 397
Argentina 477 477
Australia 550 550
Bangladesh 561 561
Bosnia Herzegovina 445 448 576 450
Brazil 354 354
Bulgaria 404 404
Canada 510 510
China 542 430 506 498 527
The Congo 435 435
Denmark 628 628
Germany 533 506 500 494 508
Estonia 375 375
The Philippines 509 498 509
France 441 496 515 457
Georgia 438 438
Greece 477 448 473
Hungary 540 540
India 567 496 530 504 552
Italy 505 450 416 496 449 456
Croatia 437 437
Lebanon 477 477
Libya 565 565
Cross-Nationally Comparative Research on Racial and Ethnic Skill…

Lithuania 370 370


Macedonia 391 391
Morocco 439 439
The Netherlands 500 511 511 508
(continued)
1187
Table 27.1 (continued)
1188

Countries of destination
Countries of origin AU AT BE CH DE DK EL IE LV LU NL NZ SC Total
New Zealand 501 535 501
Nigeria 465 465
Ukraine 476 476
Pakistan 448 532 475 462
Poland 513 473 480 556 481
A. Dicks et al.

Portugal 455 417 427


Rumania 433 614 442
Russia 431 537 490 474
Serbia Montenegro 440 426 447 432
Slovenia 508 508
Slovakia 490 490
Spain 455 535 456
Czech Republic 495 495
Turkey 395 417 420 399 415 471 430
United Kingdom 539 514 562 554 542
United States 531 531
Vietnam 557 543 557
Belarus 493 493
Zimbabwe 613 613
South-Africa 517 552 550
Sweden 614 614
Mean (immigrants) 524 430 443 436 428 432 426 512 487 422 474 543 542 468
Mean (natives) 526 509 525 509 520 495 472 518 494 501 528 529 513 513
Mean (total) 526 500 518 494 511 494 468 518 494 484 522 531 516 508
Source: PISA, 2003, own computations; AU Australia, AT Austria, BE Belgium, CH Switzerland, DE Germany, DK Denmark, EL Greece, IE
Ireland, LV Latvia, LU Luxembourg, NL The Netherlands, NZ New Zealand, SC Scotland
Cross-Nationally Comparative Research on Racial and Ethnic Skill… 1189

countries Canada, Chile, France, Hungary, Iceland, Japan, Poland, Spain,


Sweden, the UK, and the US do not differentiate countries of origin in any
wave of PISA. Others, like Norway, only differentiate between few origin
countries. In PIRLS and TIMSS, differentiation of origin countries is not
possible either. In some instances, basic information on whether the child
and/or the parents were born in the test country is lacking as well. These
countries cannot be part of cross-national analyses of origin differences, which
limits the number of countries that can be analyzed. As we will show later, this
has serious consequences for the strength of hypotheses testing.
Second, practical problems arise even if countries do permit asking about
test takers’ countries of origin. Questions asked to respondents about origin
countries and, particularly, what answers they can give are often subject to
national considerations. Not all immigrant groups are present in all countries,
and origin groups are usually small. This also has consequences for statistical
models. Different studies deal with this problem in different ways, and while
we will revisit these important drawbacks later in this chapter, for now, it is
good to keep the caveats in mind while evaluating the research findings we
will discuss.

Education Systems in Cross-National Perspective

One important explanation for cross-national variation in immigrant pupils’


performance is the educational system. When researching the role of educa-
tion systems on immigrant children’s educational performance, cross-national
assessment data offer an obvious advantage over single country studies: cross-­
national designs allow for directly comparing the educational systems of dif-
ferent countries. If the cross-national variance shown in Table 27.1 can be
explained by the properties of educational systems, this provides strong evi-
dence for the relevance of the educational system for immigrant children’s
educational performance.
One necessary ingredient for such research is the availability of cross-­
nationally comparable measures of relevant education system qualities. Here,
researchers have recently made important progress. For example, Bol and van
de Werfhorst (2011) constructed and validated cross-nationally comparable
measures of educational systems’ levels of differentiation, standardization,
vocational orientation and vocational specificity. Garrouste (2010) provides a
useful data base of policy reforms, and Braga et al. (2013) have constructed
some valuable measures that are comparable between countries and over time.
The OECD provides numerous aggregate measures that can readily be used as
1190 A. Dicks et al.

contextual characteristic (OECD 2013a, b). An overview of cross-national


variation in education systems between some relevant countries is presented
in Table 27.2.
The most important drawbacks in this research line stem from unin-
tended consequences of using cross-nationally comparable measures of
education system characteristics. First, while the construction of cross-
nationally comparable measures represents a significant improvement,
their validity is not above debate. While achieving a high level of cross-
national comparability, comparative measures obfuscate national idiosyn-
crasies of system characteristics almost by definition. For example,
work-place based vocational education systems differ widely between
countries, both in the way they are construed and how they are put into
practice. By reducing the complexities of work-­place based vocational edu-
cation to a single measure “vocational specificity”, we might misspecify the
relation it has with educational outcomes. This is an important caveat that
is often overlooked.
Second, this research line directs focus to the system characteristics for
which measurements have been constructed. Research has well-established
that differentiation and standardization are relevant for educational perfor-
mance and inequality (Van De Werfhorst and Mijs 2010). However, particu-
larly for immigrant children, different education system traits (e.g.
inclusiveness, or the ability to deal with immigrant children’s specific needs)
might be relevant. Third, and related, educational system research focuses on
national-level policies and system traits, while mostly disregarding within-­
country differences in policies.
To give one recent example, Levels et al. (2017) analyze data from PIAAC
(OECD Programme for the International Assessment of Adult Competencies)
and show that the performance gap between 1.5 generation migrants and
natives strongly correlates with education system traits. Using various aggre-
gate measures constructed from PISA (i.e. percentage of immigrant students,
percentage of schools with a majority of immigrant children, average math
gap), they suggest that 1.5 generation immigrant children do better in educa-
tion systems that are better equipped for dealing with the specific needs of
immigrant children. However, the way in which education is equipped for
immigrants’ special needs varies widely between countries, and this variation
is masked by using these aggregate measures. So, while these analyses point
towards a plausible explanation for cross-national performance differences,
specific country case studies need to be conducted to further dissect the
explanatory mechanisms.
Table 27.2 Cross-national variation in education system characteristics
% pupils in Immigrant Horizontal Horizontal
% of schools with student stratification stratification
Immigrant many gradient on Vertical (between (within Vocational Vocational
students immigrants math stratification schools) schools) specificity CE orientation Standardization
Austria 16 22 −45 0.068 2.234 −1.368 32.7 0.000 1.665 −1.396
Belgium 15 21 −29 1.005 0.820 −0.780 3.3 0.000 1.020 1.079
Canada 29 43 −9 0.376 −0.641 0.615 0 0.510 −1.703 −0.054
Cyprus – – – – – – – – – –
Czech 3 1 −24 −0.127 1.000 −1.547 35.5 0.000 1.841 −0.715
Republic
Denmark 9 8 −60 −0.217 −0.867 −1.071 47.7 1.000 0.531 −1.517
Estonia 8 13 −15 −0.537 −0.657 −0.416 30.9 1.000 −0.389 −
Finland 3 1 −80 −0.590 −0.983 −1.063 10.5 1.000 0.811 0.338
France – – – 0.929 −0.031 −0.620 11.3 1.000 0.456 1.217
Germany 13 17 −32 0.431 0.522 0.060 45 0.440 0.967 −1.244
Ireland 10 7 4 0.278 −0.398 1.607 3.8 1.000 −1.032 −0.329
Italy 7 6 −31 −0.065 0.780 −0.460 0 1.000 0.215 1.476
Netherlands 11 12 −34 0.542 1.216 1.323 20 1.000 1.350 −0.092
Norway 9 7 −46 −0.885 −0.953 −0.768 13.3 1.000 0.966 –
Spain 10 10 −45 0.751 −0.930 0.932 2.8 0.000 0.063 −0.578
Sweden 15 19 −57 −0.488 −0.879 1.448 0 0.000 0.765 −1.396
United 13 16 −9 −0.644 −0.732 1.816 0 1.000 0.516 –
Kingdom
United 21 34 −5 0.843 −0.675 −0.218 0 0.090 −1.823 −0.559
States
Sources: PISA 2012, Bol and van der Werfhorst (2011)
1192 A. Dicks et al.

Social Policy in Cross-National Perspective

Over and above the education system, immigrant children’s “context of recep-
tion” is – in theory – determined by a wide variety of country-level contextual
conditions, such as:

• Anti-discrimination legislation
• Welfare arrangements
• Labor market arrangements
• Immigration regulation
• Integration policies
• Political climate
• Economic development
• Societal openness
• Ethnic diversity
• Democratic history

Cross-national designs are well-suited for studying the effects of such poli-
cies and conditions at the country level. However, much of what we men-
tioned about educational systems also applies to these contextual characteristics:
they obscure country idiosyncrasies and within-country differences and shift
attention to those measures that are available.

Methods
When selecting literature for this review, we used very specific but flexible
decision rules. We imposed a number of restrictions. First, we decided to
focus completely on studies that used large-scale assessment data to study
educational performance of pupils with a migration background relative to
native pupils. More specifically, we focus on PISA, TIMMS, and PIRLS. As a
result, we study only primary and secondary education. Analyses of immi-
grant performance in other forms of education (preschool, vocational, or
higher education) fall outside the scope of this review. Secondly, we restricted
the review to studies that were truly comparative in nature, arguing that it is
the cross-national comparability of results that is the true added value of
large-scale assessment data. That is, we restricted the analyses to studies that
compare at least five destination countries. We imposed no restrictions on
scientific disciplines, but restricted the literature search to research published
in English-language between 2005 and 2017. As a further restriction, only
peer-refereed journal articles, contributions to books and edited collections,
Cross-Nationally Comparative Research on Racial and Ethnic Skill… 1193

and published working papers with at least 10 citations were considered for
analysis. These inclusion criteria guided the process of selecting papers.
However, we also sometimes considered studies that did not fulfill these crite-
ria. For example, some important studies only appeared very recently, but
may contribute significantly to the research tradition.
We sampled specific papers in the following process. First, we searched the
Core collection database of Web of Science using the advanced search func-
tion to incorporate Boolean arguments. We searched for the topical keywords
“*migra*”, to capture all spellings of related terms such as “Immigrants” or
“Migration”, and combined this with either “PISA”, “PIRLS”, or “TIMSS”.
This search yielded 90 results. We then manually refined the results to the five
most populated fields “Education Educational Research”, “Economics”,
“Demography”, “Sociology”, and “Psychology educational” (61 results). After
restricting the list to articles that have been cited at least 10 times and that
addressed the issue of immigrant pupils in a cross-national design, we ended
up with a short-list of five articles.
To give our search a broader scope, we also consulted the database of
Google Scholar using the following search query: (Migrant OR Immigrant
OR Immigration OR Migration OR Ethnicity OR Ethnic OR “First genera-
tion” OR “Race” OR “Second generation”) AND (PISA OR TIMSS OR
PIRLS). This search yielded >100,000 results. To better manage search results,
we continued our research not by means of the Google Scholar webpage, but
with the assistance of Publish or Perish Software (Harzing 2007). Publish or
Perish goes through Google Scholar’s database and yields the 1000 most rel-
evant search hits (i.e., the first 1000 hits as listed by Google Scholar). We once
again restricted the list to relevant works that were cited at least 10 times,
however we did not discriminate by scientific discipline. This yielded a list of
19 results. All articles found via Web of Science were also found via Google
Scholar. In addition, we included four relevant studies which we did not find
by the above described search routine. In three cases, this was because they
were cited less than ten times. The full list of selected works is shown in
Appendix B along with additional information on the type of analyses and
data sets.

 Review of the Literature: Four Research


A
Questions
The literature has evolved in four main ways, corresponding with four main
research questions. Making use of the first PISA waves, researchers first started
to quantify cross-national differences in immigrant performance and assessed
1194 A. Dicks et al.

cross-national variability in the strength of predictors at the levels of (1) indi-


viduals and (2) schools. Almost simultaneously, they also started to assess the
role of macro-level contexts of (3) destination countries, mostly using multi-
level regression techniques to account for cross-national variability. With the
publication of PISA 2003 it became possible to also assess origin differences.
Papers that (4) account for origin differences, quantify these differences, or try
to explain these differences are the fourth main area of research. We discuss
the main findings from these traditions below.

Individual Level Predictors of Educational Success


of Immigrant Children

Really the first seminal work in this tradition was the publication of Marks
(2005). Using PISA 2000 data, Marks (2005) analyzed reading and mathe-
matical literacy of first- and second-generation immigrants in a large number
of countries. He concluded that in most countries social-economic, social-­
cultural and school characteristics explain the better part of the difference in
educational achievement between native and immigrant pupils. Also, the pro-
ficiency in the destination language is affecting the educational attainment of
immigrants. Nevertheless, Marks found international differences in the way
these determinants affect the educational performance of immigrants. Only
in Belgium, France, Portugal, the United Kingdom, and the United States,
differences in reading scores could fully be explained by these determinants.
In Austria, Denmark, France, the Netherlands, Sweden, Switzerland, and the
United Kingdom, reading literacy scores of second-generation immigrants
proficient in the destination language remain substantially lower than those of
comparable natives. For mathematical literacy, similar results were found: in
Austria, Denmark, France, Netherlands, Sweden and Switzerland, second-­
generation immigrant pupils who speak the language of their destination
country on average score 20 points lower on the mathematical proficiency
scale. Again, effects vary substantially per destination country. Suggested
explanations for these findings are socio-economic, sociocultural and school
factors. Of these, socio-economic factors are assessed to be most important.
Another early contribution was given by Schnepf (2006), who analyzed
differences in mathematical literacy between first- and second-generation
immigrants and native pupils in ten Western countries, using PISA, TIMMS
and PIRLS. In general, first-generation migrants proved to be less mathemati-
cally literate than second-generation migrants, who in turn were less mathe-
matical literate than natives were. Both findings were explained by the
Cross-Nationally Comparative Research on Racial and Ethnic Skill… 1195

influence of time: the longer immigrants stayed in their country of destina-


tion, the better they would perform at school. In addition, the overall lower
educational achievement of immigrants can be explained through micro-level
characteristics such as socio-economic background and proficiency in the des-
tination language. Results are controlled for the influence of school character-
istics, such as the levels of socio-economic and ethnic segregation; the latter
proves to be of some importance for explaining different levels of mathemati-
cal literacy between natives and second-generation immigrants. However,
these explanations do not solve the whole puzzle. The found effects differ
strongly per country of destination, in strength as well as in direction. Schnepf
(2006) suggests that these differences may be explained by selection mecha-
nisms in the migration process.
Many of the theoretical puzzles to be solved later were already evident from
the early papers. Before moving to describing these, let us first review the
main substantive conclusions on the relevance of individual-level predictors.
First, the body of research confirms that many of the individual and school-­
level predictors of educational success of native children are also important for
immigrant children. For example, socioeconomic background explains most
of the immigrant-native gap in educational achievement. However between
countries, there is variation in the importance of socioeconomic background.
In the US, socioeconomic background completely explains the immigrant-­
native gap (Marks 2005). In Austria, France, Germany, and Luxembourg it
explains about half of it (Marks 2005; Schnepf 2006, 2007). In Belgium,
Denmark, Greece, the Netherlands, Norway, and Switzerland the decline is
somewhat smaller but still substantial (Marks 2005; Schnepf 2006, 2007).
Second, a number of variables explicitly related to migration are important
as well. First, generation status and length of stay are important. This is in line
with classic sociological assimilation theory. According to the classical defini-
tion, assimilation is “a process of interpretation and fusion in which persons
and groups acquire the memories, sentiments, and attitudes of other persons
and groups and, by sharing their experience and history, are incorporated
with them in a common cultural life” (Park and Burgess [1921] 1969: 735).
From a micro-level perspective, assimilation means that individual migrants
will gradually assimilate to the cultural patterns of their destination countries.
When initially creating the theoretical concept, life-cycle effects were thought
to be influencing the assimilation process; later, the notion of cohort-effects
was added to the framework (Warner and Srole 1945). It was theorized that
within time, first-generation migrants would adopt cultural patterns of their
host societies. Also, second-generation migrants would be better assimilated
into societies than their parents.
1196 A. Dicks et al.

In addition to Marks (2005) and Schnepf (2006), several empirical find-


ings support this theory. Entorf and Minoiu (2005) compared the effects of
migration background in different countries. In five of the nine countries they
analyzed, students profited from being born in the destination country. The
effect of a parental migration background, however, differed widely between
countries. In some cases, having immigrant parents posed a penalty (in FR,
DE, and UK) in other cases having immigrant parents delivered a premium
(in AUS and CN). Not only do immigrant students often perform worse than
natives, the dispersion of achievement scores is also greater for immigrants
than for natives (Schnepf 2008). In countries like Germany and Switzerland,
achievement inequality is especially driven by very low achieving immigrants.
In Canada and Australia, the mechanism is a different one. Here, relatively
low-achieving immigrant pupils can perform better than their native counter-
parts (Schnepf 2008).
Time of stay is also relevant, as assimilation theory would predict. For
first-­generation immigrant children, the age of arrival is negatively related to
their achievement (Cobb-Clark et al. 2012; Shapira 2012).But other factors
are important too. An important predictor is L2 language proficiency. Good
command of the national language is imperative for educational success.
Speaking the destination country’s language at home is a widely-used proxy
for students’ and parents’ language skills. Indeed, studies consistently find a
substantial positive effect of speaking the national language at home (Cobb-­
Clark et al. 2012; Dronkers et al. 2012; Dustmann et al. 2012; Entorf and
Minoiu 2005; Hillmert 2013; Marks 2005; Ruhose and Schwerdt 2016;
Schneeweis 2011; Schnepf 2007, 2008; Shapira 2012). Significant cross-­
national differences occur. The effect is relatively large in Germany and
Finland (Entorf and Minoiu 2005). In France, the migrant population is
mostly made up of immigrants from former colonies in Northern Africa,
whose populations to a large extent have fluent command of French.
Consequently, language spoken at home has a smaller effect than, for exam-
ple, in neighboring Germany (Entorf and Minoiu 2005). In English-speaking
countries, immigrant students are generally achieving similar to – or even
better than – natives (Dustmann et al. 2012; Schneeweis 2011). In
Anglophone countries, language contributes less to the explanation of immi-
grant-native gaps than in other continental European countries (Entorf and
Minoiu 2005). In the UK, migrants mostly come from India or Ireland.
Those who do come from non-English-speaking countries are mostly
Europeans. Australia shows a similar pattern (Entorf and Minoiu 2005). The
exception to the rule is New Zealand (Entorf and Minoiu 2005; Marks 2005;
Schnepf 2007).
Cross-Nationally Comparative Research on Racial and Ethnic Skill… 1197

What more general lessons can we draw from this? It is worth noting that
the papers we examined paint a highly consistent picture about the relevance
of family backgrounds and individual-level traits for immigrant children’s
educational performance. However, it should also be noted that the effects of
individual-level traits vary in strength between countries. This will prove an
interesting source of variance, to be explained by destination country traits.
Furthermore, the findings on the relevance of language proficiency also point
toward the relevance of origin countries.

 chool-Level Predictors of Immigrant Children’s


S
Educational Success

Classic sociology of education maintains the socioeconomic composition of


schools is seminal in predicting both the level of performance, and the socio-
economic gradient on performance. Indeed, the average sociocultural status
of a school’s parents has a strong positive effect on the achievement of both
immigrant and native pupils (Dronkers and Van der Velden 2013). Ethnic
diversity has a negative impact on immigrant pupils’ achievement – indepen-
dent of the destination country (Dronkers and Van der Velden 2013). The
origin country of peers matters as well. For instance, a higher share of peers
with a non-Islamic Asian background increases achievement, irrespective of
pupils’ background (Dronkers and Van der Velden 2013).
When taking into account that the effects of socioeconomic and ethnic
school composition on immigrant students’ achievement are cross-nationally
diverse, some interesting country specific mechanisms come to light. In
Australia and Canada, higher immigrant proportions in school increase immi-
grant pupils’ achievement (Schnepf 2007). In the Netherlands, Sweden, UK,
and US, the immigrant share is not important for achievement after control-
ling for other factors (Schnepf 2007). In Switzerland, Germany, New Zealand,
and France, immigrant the share reduces achievement of immigrant pupils
(Dronkers and Levels 2007; Schnepf 2007). On average, however, the effects
of ethnic segregation are miniscule compared to the effects of socioeconomic
segregation (Dronkers and Levels 2007).
Another aspect of schools is the pupil-teacher ratio. Immigrant pupils
might need special attention from teachers. The pupil-teacher ratio quantifies
the level of attention a teacher can give to each pupil. Empirically, fewer pupils
per teacher increase the achievement of pupils (Dronkers and Van der Velden
2013). Yet, this effect is small and does not differ between immigrant and
native pupils. A similar measure is teacher shortage, which does not only
1198 A. Dicks et al.

relate to the quantity of teachers, but also the quality. Dronkers and De Heus
(2016) show that in schools with a lack of qualified teachers, immigrant pupils
perform worse.
Nevertheless, the achievement level of peers and the school composition
only account for a small part of the immigrant-native gap. Whereas in Nordic
countries, peer quality and school characteristics account for 10% of the
immigrant-native gap, they only account for 1% in Central European coun-
tries (Dustmann et al. 2012).

Destination Countries: Contexts of Reception

Researchers were quick to recognize the value of assessment data for analyzing
and understanding the relevance of contextual characteristics of destination
countries. This cross-national testing has led to important theoretical insights.
For example, cross-national tests confronted assimilation theory with poten-
tially anomalous findings. If time is the only element relevant for immigrant
integration, immigrants from different countries of origin should assimilate at
about the same rate into the societies of their destination countries. These
expectations were found to be false: macro-level differences were found in all
dimensions of immigrants’ integration. Several research findings indicate that
macro-level differences also play a role in the educational performance of
immigrants.
In line with the theoretical importance of the context of reception for
explaining immigrant integration, various cultural, structural and institu-
tional contextual characteristics have been examined. For example, the share
of the population with an immigration background appears negatively related
to the immigrant-native gap (Cobb-Clark et al. 2012). However, the larger
origin groups in destination countries are, the better immigrant children
belonging to the groups perform at math (Levels et al. 2008). Also, socioeco-
nomic composition of immigrant communities in destination countries
appears to matter as well: the better the average socioeconomic position of
origin groups compared to natives, the higher children’s performance (Levels
et al. 2008).
Institutional characteristics related to educational systems, immigration
laws and social policies are also important. For example, non-comprehensive
school systems increase achievement inequality between immigrant and native
students (Cobb-Clark et al. 2012; Entorf and Lauk 2008). This can partly be
explained by different levels of peer interactions between high and low ability
students (Entorf and Lauk 2008). However, some ability tracking per subject
can also reduce the immigrant-native gap (Cobb-Clark et al. 2012).
Cross-Nationally Comparative Research on Racial and Ethnic Skill… 1199

Apart from the differentiation between educational systems, the age of


selection into tracks seems crucial as well. Causal evidence points to an over-
estimation of the negative effects of early tracking for immigrant pupils
(Ruhose and Schwerdt 2016). However, early tracking substantially disadvan-
tages immigrant children who do not speak the national language at home
(Ruhose and Schwerdt 2016). Similarly, in countries that allow for grade
retention, immigrant children are more likely to repeat grades and the
immigrant-­native achievement gap is larger (Park and Sandefur 2010).
The importance of time spent in school quantifies the relation of the edu-
cational system and the role of the parents for immigrant children’s integra-
tion. In (pre-)school, pupils can learn from each other and become acquainted
with the culture and the language of the country (cf. Crul and Vermeulen
2003). In countries with longer and more schooling days, pupils are also less
dependent on the support of their parents, who were often not educated in
the destination country. Because the level of support immigrant children
receive will be lower than the support native parents can offer, time spent in
school can decrease the immigrant-native gap (see Alba et al. 2011).
Empirically, several mechanisms have been tested in a cross-national setting.
Most findings suggest that time spent in school is beneficial for immigrant
pupils. Pre-primary education can reduce the immigrant-native gap
(Schneeweis 2011) as can a lower starting age (Cobb-Clark et al. 2012). More
instructional hours per year also predict higher relative achievement of immi-
grant children (Schneeweis 2011). However, effects depend on outcome (e.g.
math or science) and model specifications.
Education systems are not the only institutional trait that may be relevant
for immigrant children’s performance. Immigration laws also appear to mat-
ter. In traditional immigration countries like Australia, Canada, USA, and
some extent in the UK, immigrant children are often less disadvantaged than
in Continental Europe. This can partly be explained by selective immigration
policies of these countries (Levels et al. 2008). This is a selection effect: because
parental education is highly correlated with the education of their children,
and immigrants in these countries are relatively highly educated, their chil-
dren achieve better (Dustmann et al. 2012).
Another potentially relevant social policy is welfare. First generation immi-
grants appear to fare worst in countries with social-democratic welfare regimes
than in liberal welfare regimes (Shapira 2012). For second generation immi-
grant students, however, a social-democratic regime may be beneficial
(Shapira 2012). Apparently, liberal regimes encourage first-generation immi-
grants to obtain skills relevant to the destination country’s labor market. On
the long term, however, social-democratic policies lend more support and
1200 A. Dicks et al.

security that enable the next generation to integrate better (Shapira 2012).
However, the effects may also be spurious. Controlling for other characteris-
tics and origin differences, Levels et al. (2008) do not find an effect of left-
wing governments.
Interestingly, destination country traits may also interact with individual-­
level and school-level predictors. For instance, the socioeconomic composi-
tion of schools has a particularly strong effect in highly stratified school
systems (Dronkers et al. 2012, Dronkers and Van der Velden 2013). The
effect of ethnic segregation also appears to vary between stratified and com-
prehensive systems and affects immigrants and native pupils differently
(Dronkers and Van der Velden 2013).

Origin Countries and Ethnicity

One important drawback of studies on destination country differences is that


they mostly do not account for differences between origin groups. This has an
obvious reason: data on immigrants’ origin is often not available. However,
given the unequal distribution of migrants from different countries of origin
over the various destination countries, an analysis of the role of education for
the performance of immigrant children without controlling for origin-effects
makes analyses vulnerable and could invite false conclusions. For a rigid test
of macro-level hypotheses, a design type that incorporates both origin and
destination effects should be used.
In an analysis of 2003 PISA data, Levels and Dronkers (2008) made use of
simple OLS regression techniques to analyze cross-national differences.
Incorporating dummy variables for 13 destination countries to measure des-
tination effects, and 15 dummy variables for regions of origin to analyze ori-
gin effects, they showed that both effects play a role in predicting educational
performance of immigrants. Building on this, Levels et al. (2008) used a so-­
called “double comparative design” (cf. Van Tubergen et al. 2004). This
enabled the analyses of multiple origin groups within multiple destination
countries. Levels et al. (2008) find that – when controlling for composition
differences, destination country differences and community effects – the per-
formance of immigrant pupils from economically less developed countries is
rather good. The same is found for pupils with a background from a politi-
cally stable origin country.
However, the relevance of economic and political circumstances may have
been overestimated because features of the educational system have not been
taken into account (Dronkers et al. 2014). Accordingly, Dronkers et al. (2014)
Cross-Nationally Comparative Research on Racial and Ethnic Skill… 1201

extended on the earlier papers and found that the performance of first genera-
tion immigrant pupils in the destination country partly depends on the length
of compulsory schooling in the origin country. In these models, economic
and political differences between origin countries do not explain the perfor-
mance of immigrant pupils.

Discussion
Although the cross-national analyses of large-scale assessment data we
reviewed have provided interesting insights in explanations of educational
performance of immigrant children in Western countries, we may identify
a number of important limitations of the papers we examined. First, data
are not without limitations. Especially PISA received much attention from
the media and in public debates. PISA is also seen as the most controversial
and most critiqued. In this section we will not focus on general criticism of
large-scale assessment data, regarding for example assessment methods and
questionnaire design (for such a review see Hopfenbeck et al. 2017). Rather,
we will focus on one specific issue related to the assessment of immigrant’s
skills using large-scale assessment tests: language. Immigrants may be at a
definite disadvantage on assessment tests because of language issues.
Constructs like mathematical ability or scientific literacy are conceptually
distinct from language skills. However, language skills are important, as
relatively low language proficiency of respondents might also hamper their
ability to, for example, understand problems on math test. In general,
observed variation in math skills may therefore partly be driven by differ-
ences in language proficiency as well. While this holds true in general, the
validity issue may be particularly problematic for assessing math skills of
immigrant children, given that for many immigrants, the language in which
the test is taken is a second language at best. On the other hand, the skills
tests in PISA purport not to measure respondents’ proficiency in skills per
se, but rather their ability to use these skills to “participate in society” as
“constructive, concerned, and reflective citizen” (OECD 2003: p. 24;107).
From that perspective, proficiency differences in math and ­numeracy as
measured in PISA are insightful even if they are driven (partly) by language
differences. TIMSS and PIRLS on the other hand specifically measure
knowledge that students attained within educational curricula relating it to
how that curriculum was intended to be taught and then implemented by
teachers (Mullis et al. 2005). Hence, the scope of TIMSS is limited to
assessing only those themes which are part of the curricula of all p
­ articipating
1202 A. Dicks et al.

c­ ountries. While this is not a limitation in PISA, the skills framework of


PISA might at times assess skills which are not part of the curricula in some
countries (Harlen 2001).
Other limitations of this literature relate to research designs. For example,
the explanation of destination country differences has important limitations
that are closely related to the common use of multilevel techniques to model
between-country variance. Multilevel analyses in principle enable the con-
struction of correct standard errors by correcting variance for clustering.
However, the merits of these analyses are limited by data limitations in at least
two ways. Firstly, analyses are not uncommonly performed on a limited num-
ber of destination countries. Country selection can be limited for several rea-
sons, such as missing origin information, administrative purposes, or limited
availability of macro data. However justified, using small numbers of destina-
tion countries limits the number of country-level characteristics that can be
considered simultaneously, and may also lead to anticonservative tests of
hypotheses (Stegmueller 2013). As a consequence, observed destination coun-
try effects may be spurious and driven by third factors not in the models, or
even statistical artefacts caused by misspecification of confidence intervals.
Destination country effects can be indicative of context effects, but need to be
corroborated by other research before being interpreted.
We should also note specific problems with analyzing origin differences
using these cross-national surveys. To quantify and understand differences
related to immigrants’ origin, researchers commonly rely on information
about the birth countries of respondents and their parents. The use of birth
countries to study origin differences should be qualified with at least four
caveats. First, country samples of origin groups in large scale assessments are
not necessarily representative for all origin groups in that country. That means
that generalizations about origin variation beyond what is observed in the
sample should be made with caution. Second, and perhaps obvious given the
fact that cross-national migration has been going on for centuries and coun-
tries’ borders hardly ever coincide with ethnic borders, birth country informa-
tion is an imperfect proxy of ethnicity. As such, origin effects should not be
equated with effects of ethnicity. Third, multilevel regression techniques are
quite commonly used to analyze small origin groups (see e.g. Levels et al.
2008). Small origin groups are to be expected, given how migration is selec-
tive, immigrant samples are often small and non-representative. However,
simulations suggest that analyzing small groups may result in biased level-2
standard errors (Maas and Hox 2005). Fourth, as our description of the evalu-
ation of origin effects suggests, the strength and statistical significance of
Cross-Nationally Comparative Research on Racial and Ethnic Skill… 1203

parameters of origin characteristics in multilevel designs can be sensitive to


model building. This means that inferring causality from these analyses should
be accompanied with the necessary caveats. So, as with destination differ-
ences, origin differences should be interpreted with caution.
Another main point of debate relates to the validity of important explan-
atory variables. Many papers point to the importance of parental education
for immigrant children’s educational success. Cross-national surveys rely
heavily on cross-nationally comparable measures of parental education.
However, the comparability of parental education between immigrants and
natives is not straightforward. Alba et al. (2011) name three aspects of this
comparison that researchers should keep in mind. First, in most countries
native parents are higher educated than immigrant parents. Second, even if
comparing immigrants to natives with comparable attainment, it remains
questionable how well the educational systems of origin and destination
countries can be compared. Third, immigrants are a selective group (e.g.
Feliciano 2005). Thus, even if immigrant parents attain comparable levels
and educational systems are in fact comparable, unmeasured traits of agency
and structure remain. However, this is the case for all analyses that draw on
these types of data, and not necessarily restricted to studies of immigrant
children.

Conclusions
With these important limitations in mind, we may draw a number of general
conclusions from this literature. First, contexts matter. Taking into account
the particularities of different social groups and national contexts is indeed
the main strength of large-scale assessment data. For example, cross-national
assessments suggest that educational tracking is likely related to achievement
differentials between immigrants and natives. Most notably, large-scale assess-
ment data corroborate the many papers suggesting that comprehensive sys-
tems give room for immigrant children to catch up with natives, whereas
non-comprehensive systems exacerbate inequalities (Entorf and Lauk 2008).
Yet, subject-wise ability tracking positively influences the achievement of
immigrant pupils Cobb-Clark et al. (2012).
However, and this is the second main conclusion, the relevance of educa-
tional systems or other contextual traits is rather humble. Contextual charac-
teristics related to origin and destination countries have only limited
explanatory value (see also Schnepf 2008). From some studies, it seems that
1204 A. Dicks et al.

origin country differences may be somewhat more important than destination


country differences. Some studies find that a higher proportion of variance is
tapped by at the origin country level than by the destination country level
(Dronkers et al. 2012). Other studies find the opposite (Levels et al. 2008).
However, in most studies, the overall contribution of both origin and destina-
tion countries to explaining variance in performance of immigrant children is
limited. To illustrate the distribution of explained variance, Fig. 27.1 shows
that of all the variation in PISA 2003 reading scores between immigrant chil-
dren, by far most variance (about 86%) is explained by individual-level char-
acteristics such as parental SES and language skills. So, while origin and
destination effects play such prominent roles in public debates on immigrant-­
native achievement gaps, the scientific evidence suggests that their impor-
tance should not be overvalued.
The third main conclusion may therefore be that individual-level predictors
are the key to understanding achievement inequalities between immigrant
and native children. The most important predictor of success remains socio-
economic status. Also, language skills are important, not only as a direct
effect, but especially in concert with other factors, e.g. high school starting age
(Cobb-Clark et al. 2012).
There is also a conclusion to be drawn on a meta-level. One of the main
hallmarks of international assessment data is that they allow for ranking coun-
tries’ based on human capital potential. Such data also allow for ranking
countries’ immigrant-native achievement gaps (cf. OECD 2013a). It is tempt-
ing to interpret the achievement gap as evidence for the extent to which gov-
ernments are successful in promoting ethnic equality in education. However,

7
4
3

Destination countries
Origin countries
Communities
Individuals
86

Fig. 27.1 Proportion of unexplained observed variance in PISA 2003 reading scores
after controlling for individual-level differences. (Source: PISA 2003)
Cross-Nationally Comparative Research on Racial and Ethnic Skill… 1205

the research we reviewed for this chapter suggests that such a straightforward
interpretation may not be merited. Countries have different immigrant popu-
lations: selective migration has amounted to an unequal distribution of immi-
grants over different receiving countries. Analyses that control for
compositional differences suggest that only a modest proportion of the total
variation in achievement gaps can be attributed to destination country differ-
ences that are not related to composition. Therefore, national league tables of
such gaps should not be interpreted as an indicator of the success with which
education systems promote equality between immigrant and non-immigrant
children. Rather, they must be viewed in their full complexity, as the result of
long-term sequences with which immigrant and native children and their par-
ents have made educational and life-decisions, within the specific contexts of
their origin and destination countries.

Future Research
Comparative studies can yield insightful conclusions about immigrant-native
achievement gaps. From these studies we can learn about plausible – indi-
vidual and contextual – explanations for cross-national differences in immi-
grant achievement. However, to understand the mechanisms at place,
quantitative analyses of large-scale data have to be supplemented with other
types of analyses. We have to rely on quasi-experimental and ethnographic
methods, in-depth case studies and specialized survey data.
For future endeavors, pooling of available data seems a promising avenue.
On the one hand, pooling survey waves can yield valuable insights regarding
trends over time. On the other hand, creative combinations of survey data can
create natural experiments (i.e. Ruhose and Schwerdt 2016). Traditional
regression-control designs rely on the assumption that all confounding factors
are observed and controlled for. Especially in multi-level designs, where the
higher level samples are often very small, this can bias the estimation. Natural
experiments can isolate causal effects by also controlling for unobservable
confounders through randomization. Causal inferences are especially relevant
to the evaluation of educational policies, and other destination country effects.
However, the search for natural experiments could potentially limit the choice
of research questions to be answered. Furthermore, researchers need to be
cautious about the possible caveats that arise when combining data from
­different assessment surveys entailing different frameworks, some of which we
have listed above.
1206 A. Dicks et al.

There remain important questions to be answered. First, there is still much


to be done examining the role of teachers. For example, teachers with an
immigration background could have different expectations and assessments
of immigrant pupils than native teachers (most of this research is related to
the US context, e.g. Dee 2005). For PISA this is partly related to the sam-
pling procedure of age groups in schools instead of classes and partly to the
absence of a dedicated teacher questionnaire. That is, until 2015 when a
teacher questionnaire was introduced in a limited set of countries (together
with the linked teacher survey TALIS, cf. Le Donné et al. 2016; OECD
2015b). For TIMSS and PIRLS, teacher information is more readily avail-
able. Nevertheless, neither PISA nor TIMSS/PIRLS gather information on
the origin of teachers.
Second, our knowledge of the (behavioral) role of the parents, above and
beyond their socioeconomic resources, is limited in comparative contexts.
Immigrant parents might have very different beliefs about the role of the
school in education which affects their involvement in schooling matters (cf.
Crozier and Davies 2007). Also here, data availability is the main hindrance
to study related mechanisms in a comparative design.
Third, the observation of destination effects should invite research aimed at
discovering possible mechanisms underlying observed macro-micro correla-
tions. For such purposes, the outcomes of cross-national analyses of assess-
ment data can serve as input for more small-scale comparative research. For
example, ethnographic research in schools in different national contexts can
provide valuable insights into the way in which national contexts lead to dif-
ferent (or similar) outcomes for immigrant children in different countries.
Paulle (2013) is an exemplary participatory ethnographic study of schools in
different national contexts, and may serve as an inspiration for scholars inter-
ested in the ethnographic study of immigrant children’s educational achieve-
ments in different national contexts.
Finally, the league tables of nations’ PISA scores has led to efforts to change
education systems in various countries (Breakspear 2012), sometimes with
the express goal of improving the chances of immigrant children (Ertl 2006;
Egelund 2008). It would be insightful to use quasi-experimental designs to
assess the causal link between these policy changes and educational achieve-
ment of immigrant children. Such studies would help to assess not only
whether these policies were effective, but also whether or not holding coun-
tries accountable to their educational output helps in improving immigrant
children’s performance.
Cross-Nationally Comparative Research on Racial and Ethnic Skill… 1207

Appendix A
PISA

The Programme for International Student Assessment (PISA) is an interna-


tional survey that “aims to evaluate education systems worldwide by testing
the skills and knowledge of 15-year-old students.” (OECD 2017). PISA is
conducted every three years, in an increasing number of participating coun-
tries. The first wave of PISA was conducted in 2000 and 2001; at that point
about in 43 countries participated. In 2015, over half a million students from
72 countries and economies participated in PISA. Students are assessed on
science literacy, mathematical literacy, reading skills, as well as collaborative
problem solving and financial literacy. The outcome variables are measured
through multiple items, directly testing the extent in which pupils are profi-
cient in certain skills. Pupils are presented with a selection of these items; item
response modelling is used to calculate plausible values on literacy. These
plausible values provide an unbiased estimate of the answers on all the literacy
items. In addition, data on schools is collected through a separate
questionnaire.

PIRLS and TIMMS

PIRLS is the Progress in International Reading Literacy Study. It runs every


five years and targets pupils in fourth grade. It assesses reading literacy
through a main survey that consists of a written reading comprehension
test, and a background questionnaire. It is implemented in about 40 coun-
tries. TIMSS is the Trends in International Mathematics and Science Study.
Since 1995, it is organized every four years in about 40–55 countries, and
focusses on pupils in fourth and eighth grades. TIMMS measures perfor-
mance in various mathematics and science domains (such algebra, geome-
try, biology), and assesses the extent to which pupils are able to solve
problems in these domains. Contextual information about schools is also
collected. Both PIRLS and TIMMS are conducted by TIMSS & PIRLS
International Study Center at Boston College and the International
Association for the Evaluation of Educational Achievement (IEA), with the
objective to allow governments to make evidence-based decisions for
improving educational policy.
Appendix B: Literature Review

Destination/
Origin
Found Countries
via Authors Year Data Method (Regions) Main result
WoS Levels et al. 2008 PISA 2003 Cross-classified 13/35 Low economic development in origin country has positive effect on
three-level immigrant pupil’s performance. Politically unstable origin
hierarchical linear countries negatively related to immigrant pupil’s performance.
model Children from immigrant communities with better socioeconomic
capital than the native population perform better
WoS Schnepf 2007 PISA 2003, PIRLS Country regressions 10 Immigrant children in English-speaking countries often fare better
2001, TIMSS than immigrant children in Continental Europe. Language skills,
1995,1999 SES, and school segregation reduce the gap between immigrants
and natives, however the gap remains in NL, DE, FR, NZ, and CH
WoS Marks 2005 PISA 2000 Country regressions 20 SES most important factor for lower performance of immigrants.
SES, sociocultural, and school factors account for 58–79% of
differences between immigrants and natives. Language spoken at
home only has minor impacts, once SES is controlled for
GS Entorf and 2005 PISA 2000 Country regressions 9 Language spoken at home is most important
Minoiu
WoS Entorf and 2008 PISA 2000 Grouped country 11(4) Non-comprehensive school systems show a larger gap between
Lauk regressions for migrant and native students. Peer effects are greater in tracked
native/migrants systems
GS Dustmann 2012 PISA 2006 (US: Pooled, Country 18(4)/7 In countries with highly educated immigrants, immigrant children
et al. PISA 2003) regressions fare better. Language spoken at home is most important. Turkish
migrants outperform children in Turkey, better school and peer
quality key determinants
WoS Levels and 2008 PISA 2003 Destination and 13/(14) Immigrants from Southern- and Central America, Northern Africa
Dronkers Origin-fixed and Western Asia have substantially lower math scores than
effects regression natives
(continued)
(continued)
Destination/
Origin
Found Countries
via Authors Year Data Method (Regions) Main result
GS Schneeweis 2011 PISA 2000/2003, Pooled, Country-­ 62(9)/(9) Time in school and early education positively related to immigrant
TIMSS group-­effects, pupil’s performance. Social and ethnic segregation are mostly
1995/1999/2003 Country unimportant
fixed-effects
GS Dronkers 2013 PISA 2006 Four-level 15/46 High ethnic diversity is especially harmful for immigrant children.
and Van hierarchical linear Pupils with a non-Islamic Asian background have an advantage
der model compared to other immigrant and to natives. Other children also
Velden benefit from presence of non-Islamic Asian pupils in school. No
effect found for socio-­cultural diversity
GS Dronkers 2007 PISA 2003 Three-level 13/(14) Ethnic segregation in school is harmful for natives and immigrants,
and hierarchical linear however for some origin groups more than for others. Still,
Levels model socioeconomic segregation is more important
GS Schnepf 2008 PISA 2003, TIMSS Quantile regression 8 Performance dispersion is greater for immigrants than for natives.
2003 (CH/DE: Dispersion is especially driven by very low performing immigrant
TIMSS 1995), pupils. Language skills are more important in lower than in higher
PIRLS 2001 performance quantiles
GS Dronkers 2012 PISA 2006 Cross-classified 15/35 Performance of migrants differs over educational systems. Future
et al. three-level research should not ignore ability grouping
hierarchical linear
model
GS Cobb-Clark 2012 PISA 2009 Country fixed-­ 34 Immigrant-native gap is larger for those who arrived at older ages
et al. effects regression and who do not speak the test language at home. Ability tracking
per subject can be beneficial for some migrant students, but full
ability tracking can be detrimental
(continued)
(continued)
Destination/
Origin
Found Countries
via Authors Year Data Method (Regions) Main result
GS Park and 2010 PISA 2000 Country fixed-­ 11 Immigrant pupils are more likely to repeat a grade. In countries
Sandefur effects regression, with grade retention, immigrant-native gap is bigger
Country
regressions,
Two-level
hierarchical linear
model
Added Dronkers 2016 PISA 2006 Cross-classified 16/35 Immigrant children’s performance suffers in countries with teacher
and de three-level shortage
Heus hierarchical linear
model
GS Borgna 2014 PISA 2006, 2009 Regression tree 17 Severe migrant-specific penalties in Western Europe. Cross-country
and ANOVA differences not attributable to origin composition. Migrant-
Contini specific and socio-economic penalties are two distinct dimensions
GS Hillmert 2013 PIRLS 2001, PISA Country regressions 5 Parental SES, test language familiarity and school context explain
2006 the immigrant-­native gap in France, and the UK. Some gaps
remain in Germany, Netherlands, and Sweden
GS Ruhose 2016 PISA, PIRLS, Diff-in-Diff 25 Cross-sectional estimates overstate effects of early tracking on
and TIMSS immigrant-native gap. However, early tracking does negatively
Schwerdt affects second-generation immigrants who do not speak the test
language at home and for first generation immigrants
GS Shapira 2012 PISA 2006 Three-level 18 First generation immigrant children perform relatively well in
hierarchical linear liberal welfare regimes and countries with standardized
model educational systems and selective immigration policies
(continued)
(continued)
Destination/
Origin
Found Countries
via Authors Year Data Method (Regions) Main result
Added Dronkers 2014 PISA 2006 Cross-classified 16/35 Higher student-teacher ratios in primary school relatively increase
et al. three-level immigrant children’s performance. For immigrant children who
hierarchical linear attended school in their country of origin at least for some time,
model the duration of compulsory education in the origin country is
positively related to their performance
Added Dronkers 2014 PISA 2009 Destination country 30/62(12) Immigrant girls have higher reading and math scores than
and fixed-effects, immigrant boys. This difference is larger among immigrants than
Kornder separately for among natives
boys and girls
Added Dronkers 2015 PISA 2009 Cross-classified 17/45 Immigrant girls have higher reading scores when the gender
and two-level equality in the origin country is higher. Gender equality is a
Kornder hierarchical linear mediator of religion
model
1212 A. Dicks et al.

Bibliography
Alba, R., Sloan, J., & Sperling, J. (2011). The Integration Imperative: The Children
of Low-Status Immigrants in the Schools of Wealthy Societies. Annual Review of
Sociology, 37, 395–415.
Borgna, C., & Contini, D. (2014). Migrant Achievement Penalties in Western
Europe: Do Educational Systems Matter? European Sociological Review, 30(5),
670–683.
Bol, T., & van de Werfhorst, H. G. (2011). Signals and Closure by Degrees: The
Education Effect Across 15 European Countries. Research in Social Stratification
and Mobility, 29(1), 119–132.
Braga, M., Checchi, D., & Meschi, E. (2013). Educational Policies in a Long-Run
Perspective. Economic Policy, 28(73), 45–100.
Breakspear, S. (2012). The Policy Impact of PISA: An Exploration of the Normative
Effects of International Benchmarking in School System Performance (OECD
Education Working Papers, No. 71). OECD Publishing.
Cobb-Clark, D. A., Sinning, M., & Stillman, S. (2012). Migrant Youths’ Educational
Achievement: The Role of Institutions. The ANNALS of the American Academy of
Political and Social Science, 643(1), 18–45.
Crozier, G., & Davies, J. (2007). Hard to Reach Parents or Hard to Reach Schools?
A Discussion of Home-School Relations, with Particular Reference to Bangladeshi
and Pakistani Parents. British Educational Research Journal, 33(3), 295–313.
Crul, M., & Vermeulen, H. (2003). The Second Generation in Europe. International
Migration Review, 37, 965–986.
Dee, T. S. (2005). A Teacher Like Me: Does Race, Ethnicity, or Gender Matter?
American Economic Review, 95(2), 158–165.
Dronkers, J. (2010). Features of Educational Systems as Factors in the Creation of
Unequal Educational Outcomes. In J. Dronkers (Ed.), Quality and Inequality of
Education. Cross-National Perspectives (pp. 163–204). Dordrecht/Heidelberg/
London/New York: Springer.
Dronkers, J., & De Heus, M. (2016). The Educational Performance of Children of
Immigrants in Sixteen Oecd Countries. In D. J. Besharov & M. H. López (Eds.),
Adjusting to a World in Motion: Trends in Global Migration and Migration Policy
(pp. 264–290). Oxford/New York: Oxford University Press.
Dronkers, J., & Kornder, N. (2014). Do Migrant Girls Perform Better than Migrant
Boys? Deviant Gender Differences Between the Reading Scores of 15-Year-Old
Children of Migrants Compared to Native Pupils. Educational Research and
Evaluation: An international Journal on Theory and Practice, 20(1), 44–66.
Dronkers, J., & Kornder, N. (2015). Can Gender Differences in Educational
Performance of 15-Year-Old Migrant Pupils Be Explained by Societal Gender
Equality in Origin and Destination Countries? Compare: A Journal of Comparative
and International Education, 45(4), 610–634.
Cross-Nationally Comparative Research on Racial and Ethnic Skill… 1213

Dronkers, J., & Levels, M. (2007). Do School Segregation and School Resources
Explain Region-of-Origin Differences in the Mathematics Achievement of
Immigrant Students. Educational Research and Evaluation, 13(5), 435–462.
Dronkers, J., & Van der Velden, R. (2013). Positive but Also Negative Effects of
Ethnic Diversity in Schools on Educational Performance? An Empirical Test Using
Pisa Data. In M. Windzio (Ed.), Integration and Inequality in Educational
Institutions. Dordrecht: Springer.
Dronkers, J., van der Velden, R., & Dunne, A. (2012). Why Are Migrant Students
Better Off in Certain Types of Educational Systems or Schools than in Others.
European Educational Research Journal, 11(1), 11–44.
Dronkers, J., Levels, M., & de Heus, M. (2014). Migrant Pupils’ Scientific
Performance: The Influence of Educational System Features of Origin and
Destination Countries. Large-scale Assessments in Education, 2(3), 3.
Dustmann, C., Frattini, T., & Lanzara, G. (2012). Educational Achievement of
Second-Generation Immigrants: An International Comparison. Economic Policy,
27(69), 143–185.
Egelund, N. (2008). The Value of International Comparative Studies of Achievement –
A Danish Perspective. Assessment in Education: Principles, Policy & Practice, 15(3),
245–251.
Entorf, H., & Lauk, M. (2008). Peer Effects, Social Multipliers and Migrants at
School: An International Comparison. Journal of Ethnic and Migration Studies,
34(4), 633–654.
Entorf, H., & Minoiu, N. (2005). What a Difference Immigration Policy Makes: A
Comparison of Pisa Scores in Europe and Traditional Countries of Immigration.
German Economic Review, 6(3), 355–376.
Ertl, H. (2006). Educational Standards and the Changing Discourse on Education:
The Reception and Consequences of the PISA Study in Germany. Oxford Review
of Education, 32(5), 619–634.
Feliciano, C. (2005). Educational Selectivity in US Immigration: How Do Immigrants
Compare to Those Left Behind? Demography, 42(1), 131–152.
Garrouste, C. (2010). 100 Years of Educational Reforms in Europe: A Contextual
Database. Munich: University Library of Munich.
Gruber, K. H. (2006). The German ‘PISA-Shock’: Some Aspects of the Extraordinary
Impact of the OECD’s PISA Study on the German Education System. In Cross-­
National Attraction in Education: Accounts from England and Germany. Oxford:
Symposium Books.
Harlen, W. (2001). The Assessment of Scientific Literacy in the Oecd/Pisa Project.
Studies in Science Education, 36(1), 79–103.
Harzing, A.-W. (2007). Publish or Perish. http://www.harzing.com/pop.htm
Heath, A. F., Rothon, C., & Kilpi, E. (2008). The Second Generation in Western
Europe: Education, Unemployment, and Occupational Attainment. Annual
Review of Sociology, 34, 211–235.
1214 A. Dicks et al.

Hillmert, S. (2013). Links Between Immigration and Social Inequality in Education:


A Comparison Among Five European Countries. Research in Social Stratification
and Mobility, 32, 7–23.
Hopfenbeck, T. N., Lenkeit, J., El Masri, Y., Cantrell, K., Ryan, J., & Baird, J.-A.
(2017). Lessons Learned from Pisa: A Systematic Review of Peer-Reviewed Articles
on the Programme for International Student Assessment. Scandinavian Journal of
Educational Research, 62(3), 333–353.
Kao, G., & Thompson, J. S. (2003). Racial and Ethnic Stratification in Educational
Achievement and Attainment. Annual Review of Sociology, 29, 417–442.
Le Donné, N., Fraser, P., & Bousquet, G. (2016). Teaching Strategies for Instructional
Quality: Insights from the Talispisa Link Data (OECD Education Working Papers).
Paris: OECD Publishing.
Levels, M., & Dronkers, J. (2008). Educational Performance of Native and Immigrant
Children from Various Countries of Origin. Ethnic and Racial Studies, 31(8),
1404–1425.
Levels, M., Dronkers, J., & Kraaykamp, G. (2008). Immigrant Children’s Educational
Achievement in Western Countries: Origin, Destination, and Community Effects
on Mathematical Performance. American Sociological Review, 73(5), 835–853.
Levels, M., Dronkers, J., & Jencks, C. (2017). Contextual Explanations for Numeracy
and Literacy Skill Disparities Between Native and Foreign-Born Adults in Western
Countries. PLoS One, 12(3), e0172087.
Maas, C. J. M., & Hox, J. J. (2005). Sufficient Sample Sizes for Multilevel Modeling.
Methodology, 1(3), 86–92.
Marks, G. N. (2005). Accounting for Immigrant Non-Immigrant Differences in
Reading and Mathematics in Twenty Countries. Ethnic and Racial Studies, 28(5),
925–946.
Mullis, I. V. S., Martin, M. O., Ruddock, G. J., O’Sullivan, C. Y., Arora, A., &
Erberber, E. (2005). Timss 2007 Assessment Framework. Boston: TIMMS & PIRLS
International Study Center.
OECD. (2003). The Pisa 2003 Assessment Framework. Paris: OECD Publishing.
OECD. (2013a). Pisa 2012 Results: Excellence Through Equity (Volume II): Giving
Every Student the Chance to Succeed. Paris: OECD Publishing.
OECD. (2013b). Pisa 2012 Results: What Makes a School Successful? (Volume Iv)
Resources Policies and Practices. Paris: OECD Publishing.
OECD. (2015a). Pisa Resuls in Focus. Paris: OECD Publishing.
OECD. (2015b). Pisa 2015 Assessment and Analytical Framework: Science, Reading,
Mathematic and Financial Literacy. Paris: OECD Publishing.
OECD. (2017). What is PISA? Online webpage at http://www.oecd.org/pisa/about-
pisa/. Accessed 1 Nov 2017.
Park, R. E., & Burgess, E. W. ([1921] 1969). Introduction to the Science of Sociology.
Chicago: University of Chicago Press Chicago.
Park, H., & Sandefur, G. (2010). Educational Gaps Between Immigrant and Native
Students in Europe: The Role of Grade. In J. Dronkers (Ed.), Quality and
Cross-Nationally Comparative Research on Racial and Ethnic Skill… 1215

Inequality of Education. Cross-National Perspectives (pp. 113–136). Dordrecht/


Heidelberg/London/New York: Springer.
Paulle, B. (2013). Toxic Schools: High-Poverty Education in New York and Amsterdam.
(Fieldwork Encounters and Discoveries). Chicago: The University of Chicago Press.
Pigozzi, M. (2006). “What Is the Quality of Education?” A UNESCO Perspective.
In K. N. Ross & I. J. Genevois (Eds.), Cross-National Studies of the Quality of
Education: Planning Their Design and Managing Their Impact (pp. 39–50). Paris:
UNESCO, International Institute for Educational Planning.
Ruhose, J., & Schwerdt, G. (2016). Does Early Educational Tracking Increase
Migrant-Native Achievement Gaps? Differences-in-Differences Evidence Across
Countries. Economics of Education Review, 52, 134–154.
Sarrazin, T. (2010). Deutschland schafft sich ab. Wie wir unser Land aufs Spiel setzen.
München: DVA.
Schleicher, A. (1999). Measuring Student Knowledge and Skills: A New Framework for
Assessment. Paris: Organisation for Economic Co-Operation and Development.
Schneeweis, N. (2011). Educational Institutions and the Integration of Migrants.
Journal of Population Economics, 24(4), 1281–1308.
Schnepf, S. V. (2006). How Different Are Immigrants? A Cross-Country and Cross-­
Survey Analysis of Educational Achievement. In Immigration and the Transformation
of Europe (pp. 200–235). Cambridge: Cambridge University Press.
Schnepf, S. V. (2007). Immigrants’ Educational Disadvantage: An Examination Across
Ten Countries and Three Surveys. Journal of Population Economics, 20(3), 527–545.
Schnepf, S. V. (2008). Inequality of Learning Amongst Immigrant Children in
Industrialised Countries (IZA Discussion Paper).
Shapira, M. (2012). An Exploration of Differences in Mathematics Attainment
Among Immigrant Pupils in 18 OECD Countries. European Educational Research
Journal, 11(1), 68–95.
Stegmueller, D. (2013). How Many Countries for Multilevel Modeling? A
Comparison of Frequentist and Bayesian Approaches. American Journal of Political
Science, 57(3), 748–761.
Van De Werfhorst, H., & Mijs, J. (2010). Achievement Inequality and the
Institutional Structure of Educational Systems: A Comparative Perspective.
Annual Review of Sociology, 36, 407–428.
Van Tubergen, F., Maas, I., & Flap, H. (2004). The Economic Incorporation of
Immigrants in 18 Western Societies: Origin, Destination, and Community
Effects. American Sociological Review, 69(5), 704–727.
Warner, W. L., & Srole, L. (1945). The Social Systems of American Ethnic Groups. New
Haven: Yale University Press.

You might also like