Free Movement of Workers and Native Demand for

Tertiary Educationa

Mirjam Bächli Teodora Tsankova

Downloaded from by guest on September 4, 2025. Copyright 2023

January 31, 2023

Abstract

We investigate how natives’ demand for tertiary education responds to a reform that

relaxed employment restrictions for foreign workers in Switzerland. The policy change

led to a sharp increase in cross-border commuting into Swiss areas close to the border.

We find that enrollment in universities with a focus on applied studies rises in affected

commuting zones, and this increase is driven by natives with a vocational background.

Enrollment rises mostly in non-STEM fields that foster skills less transferable across

countries. We show that enrollment results are consistent with a rising wage premium

for tertiary non-STEM degrees.

JEL Codes: F22, I26, J24, J61, R23

a
Mirjam Bächli is a Postdoctoral Researcher in Economics at the University of Lausanne (mir-
[email protected]). Teodora Tsankova is an Assistant Professor of Economics at Tilburg University
([email protected]). We are grateful to Andreas Beerli, Michel Beine, Beatrix Eugster,
James Fenske, Reto Föllmi, Albrecht Glitz, Clément Imbert, Michael Knaus, Rafael Lalive, Joan Llull,
Sharun Mukand, Panu Poutvaara, David Schindler, Michael Siegenthaler, Stefan Wolter, Ulf Zölitz, Josef
Zweimüller and two anonymous referees for comments and suggestions. We thank seminar participants at
Tilburg University, University of St.Gallen, and University of Warwick for helpful comments. This paper
uses confidential data from different Swiss authorities. The authors are willing to assist in obtaining the data.
Each author declares that they have no relevant or material financial interests that relate to the research
described in this paper. An Online Appendix is included.

doi:10.3368/jhr.0721-11805R2 1
This open access article is distributed under the terms of the CC-BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0) and is freely available online at: http://jhr.uwpress.org
 1 Introduction

Higher education has gained momentum in the developed world with one in three people
in the OECD holding a tertiary degree in 2020 (OECD, 2022). Schooling decisions have a
significant impact on individual outcomes as there are substantial returns to acquiring higher
education. For example, OECD graduates with a tertiary degree earned, on average, 55%
more than those with an upper-secondary degree. Study decisions are likely tied to labor
market opportunities, which depend on the relative skill distribution. International mobility
can significantly alter the skill distribution in the host countries. Thus, if immigration
Downloaded from by guest on September 4, 2025. Copyright 2023

reforms result in a substantial inflow of workers of a particular skill group, they can change
incentives for natives to demand schooling.
In this paper, we focus on an inflow of skilled foreign workers who could either encourage
or discourage natives to enroll into tertiary education depending on how labor market out-
comes are affected. The traditional view is that skill groups most affected by an immigrant
inflow face worse labor market conditions (Borjas, 1995; Borjas and Doran, 2012; Dustmann
et al., 2012). This suggests that an inflow of skilled workers is likely to weaken native in-
centives to accumulate human capital. At the same time, there is evidence that skilled
immigrants boost total factor productivity and innovation (Moser et al., 2014; Peri et al.,
2015; Hunt, 2017), which suggests that incentives to invest in human capital may increase.
This positive effect is likely to be particularly strong for skills that are complementary to
the ones brought by foreign workers (Peri and Sparber, 2011; Ottaviano and Peri, 2012).
We explore how changes in labor market conditions induced by a free movement reform
affect natives’ educational decisions at the tertiary level. Switzerland offers a unique context
to study this research question. Specifically, the Agreement on the Free Movement of Persons
(AFMP) abolished restrictions to access the Swiss labor market for foreign workers from
the European Union (EU) and European Free Trade Association (EFTA), including cross-
border commuters. Commuters are employed in Switzerland but reside abroad, typically
in neighboring countries. With the reform’s introduction, the number of frontier workers

2
 permanently increased. Moreover, the Swiss education system enables us to isolate education
demand from supply forces as having completed upper-secondary education with a matura
exam guarantees admission. Similar to other Western European countries, Switzerland’s
dual education system gives access to tertiary education at universities (UNIs) to graduates
with general training and at universities of applied sciences (UAS) to graduates with general
and vocational training. Different educational backgrounds at the upper-secondary level
are linked to a different level of labor market experience and are likely to lead to different
enrollment decisions in response to changes in labor market conditions. Finally, we have
access to administrative data on all individuals enrolled in tertiary education, which allows
Downloaded from by guest on September 4, 2025. Copyright 2023

us to precisely quantify demand by institutional type and study field.

Our empirical strategy combines the timing of the AFMP implementation with cross-
sectional variation in distance to the Swiss border in a difference-in-differences framework.
Motivated by the fact that commuting costs rise with distance, we define Swiss areas close to
the international border as affected commuting zones and those further away as non-affected
(Dustmann et al., 2017; Beerli et al., 2021). Indeed, approximately 90% of cross-border
commuters are employed within 30 minutes of travel time from the border in 2016. We
assign native students to their place of residence when they took their matura exam under
the assumption that individuals consider local information at the time of enrollment. There
is no evidence suggesting that trends in native educational and labor market outcomes would
have been different in affected and non-affected commuting zones absent the reform.
We find that the share of commuters in affected commuting zones grew by 3.3 percentage
points relative to the non-affected areas in the post-reform period. This effect corresponds
to 22.9% relative to the average exposure in the pre-reform period, and it is driven by
skilled commuters with an upper-secondary or tertiary education. We find that native first
year enrollment at UAS rises in the post-reform period in affected commuting zones by 1.1
percentage points, which is 14.1% relative to the pre-reform average. UNI enrollment in
affected relative to non-affected commuting zones does not change due to the reform.

3
 We then map occupations to fields of study using survey data and classify the fields
according to the extent to which they are affected by the exposure to commuters. Subjects
are considered to be affected if they are linked to occupations that cross-border commuters
hold relatively more often than resident workers. We find that enrollment in less affected
fields of study at UAS rises in the post-reform period in commuting zones close to the
border. These are mostly non-STEM fields that typically require more country-specific skills
compared to STEM fields. Our findings are robust to different treatment definitions, outcome
measures, and control variables.
The reform directly affected the workforce composition by raising the share of skilled
Downloaded from by guest on September 4, 2025. Copyright 2023

foreign workers. We document a rise in the wages of natives with tertiary education and the
likelihood that they hold a managerial position, while wages decrease for those with an upper-
secondary degree. Furthermore, we investigate whether these effects differ by occupation.
Wages for native tertiary-educated workers in affected areas increase for STEM and non-
STEM workers, and the share of employed in management rises in particular for the latter
group. Wages at the upper-secondary level increase for STEM workers and decrease for
non-STEM workers. These results suggest complementarities between foreign workers, who
are overrepresented in STEM professions, and high-skilled natives employed in non-STEM
jobs.
We next look at the educational background of first-year tertiary students, which can be
linked to their awareness of labor market conditions. We find that the rise in enrollment
at UAS is driven by individuals with a vocational background at the upper-secondary level
who are prepared to enter the labor market. This gives them knowledge of local labor
market conditions, which is in contrast to individuals with a general background who are
prepared to enter tertiary education only. More specifically, we show that the increase in
enrollment comes from natives with a non-STEM background implying that natives respond
by advancing their non-STEM skills, which is consistent with the rising labor market returns
to non-STEM degrees that we estimate.

4
 We contribute to the literature that links native educational outcomes to immigration,
which has so far relied on evidence from the United States. Early work finds a negative effect
of immigration on high school graduation rates of American-born minorities and argues that
it is likely driven by competition for school resources (Betts, 1998). More recently, Hunt
(2017) differentiates between adult immigrants and immigrants of school age. The results
show that a higher share of low-skilled adult immigrants has a positive impact on high school
completion through its effect on labor market conditions, while a higher share of school-aged
immigrants has no effect. McHenry (2015) documents that low-skilled immigration has led
to a rise in native educational outcomes at the secondary and post-secondary level. Llull
Downloaded from by guest on September 4, 2025. Copyright 2023

(2018) argues that the direction of response varies across the native population depending on
individual labor market returns to education. Most of the existing work either assumes an
exogenous immigrant allocation or uses a shift-share instrumental variable strategy, which
relies on strong assumptions (for example, Goldsmith-Pinkham et al., 2020). In contrast,
we focus on a policy experiment as an exogenous source of variation. The inflow of foreign
workers we explore consists of cross-border commuters who do not compete with natives for
school resources. This helps us to capture solely variation in local labor market conditions.
Our mapping between occupations and fields of study contributes to the literature on
differences in occupational choices between immigrants and natives. Studies document that
foreign-born workers are more often employed in scientific and technical occupations than
natives (Peri and Sparber, 2009; Hunt and Gauthier-Loiselle, 2010; Peri and Sparber, 2011;
Hanson and Slaughter, 2018). We confirm these findings in a context where the foreign
workers are culturally and linguistically similar to the natives. Few studies link immigrant
occupational choices to native enrollment in specific study fields. Ransom and Winters
(2021) find that an increase in foreign STEM workers drives natives, specifically black male
students, away from STEM fields of study in the United States. Cortés and Pan (2015)
document a similar crowding-out effect from nursing studies. We add to this literature by
considering all study fields, which increases generalizability. Grouping fields by the intensity

5
 of expected labor market competition with foreign workers enables us to link enrollment
decisions to labor market conditions at the field level.
The education literature shows that expected earnings and employment prospects matter
in the study field choice (Beffy et al., 2012; Long et al., 2015; Wiswall and Zafar, 2015;
Schweri and Hartog, 2017; Acton, 2021; Abramitzky et al., forthcoming). A number of
related studies exploit business cycles to evaluate the impact of opportunity costs on demand
for education. There is evidence that enrollment is countercyclical at lower educational levels
(Ayllón and Nollenberger, 2021), in college (Dellas and Sakellaris, 2003; Long, 2014), and in
graduate school for women (Johnson, 2013). In comparison, we use an immigration reform
Downloaded from by guest on September 4, 2025. Copyright 2023

that creates exogenous variation in local labor market conditions. We distinguish between
individuals with general and vocational backgrounds to identify the groups of individuals
most responsive to the changes in labor market conditions. This level of detail is novel to
the literature and allows us to link responsiveness to reform changes with experience in the
labor market.
To understand drivers of enrollment decisions, we investigate the labor market effects of
a free movement of workers reform. There is mixed evidence on the impact of an inflow of
foreigners on native labor market outcomes (Borjas, 2003; Ottaviano and Peri, 2012; Dust-
mann et al., 2016). Unlike most of the existing literature that looks at resident immigrants,
we focus on cross-border commuters. In an early study, Dustmann et al. (2017) examine a
temporary increase in low-skilled Czech frontier workers into Germany after the Berlin wall
fell. The authors find a decline in wages and an even stronger drop in employment out-
comes for natives. Looking at the same reform as we do, Beerli et al. (2021) find a positive
effect on the wages of high-skilled natives and in addition document a positive impact on
labor demand, productivity, and innovation. Cristelli and Lissoni (2020) show that natives
who collaborate with cross-border inventors benefit from higher productivity. We extend
this literature by examining the reform effects on native human capital accumulation, which
likely have a long-run impact on the native skill composition. It is important to consider

6
 such adjustments to understand the far-reaching effects of immigration on host country labor
markets.
The remainder of the article is organized as follows. In Section 2 we discuss the regulatory
framework applied to cross-border commuters and the educational system in Switzerland.
In Section 3 we describe the data and outline the empirical strategy. In Section 4 we present
our results on commuter exposure and enrollment outcomes and discuss the mechanisms in
Section 5. In Section 6 we conclude.

2 Context
Downloaded from by guest on September 4, 2025. Copyright 2023

2.1 Cross-Border Commuting

Individuals with a citizenship from an EU or EFTA member state working in Switzerland are
subject to the rules outlined in the Agreement on the Free Movement of Persons (AFMP),
which was signed in June 1999, approved by the electorate in May 2000, and introduced on
June 1, 2002.1 The agreement affects all workers from EU and EFTA countries. We focus on
cross border commuters who are non-Swiss by nationality and require a working permit to
be employed in Switzerland. Because they need a working contract from a Swiss employer
to receive or extend such a permit, frontier workers are employed individuals.
Before the AFMP, cross-border commuters and the firms that wanted to hire them had
to fulfill several requirements. Commuters had to reside in formal border zones in the
neighboring countries and were only allowed to work in similarly defined zones in the border
area of Switzerland. Permits were tied to a specific employer and valid for up to one year,
after which they had to be renewed. Commuters had to return to their place of residence
daily. Furthermore, employers had to prove that the vacancy could not be filled by a native
1
The AFMP is a bilateral agreement. All restrictions that limited the free movement of Swiss nationals
to EU and EFTA member states were completely removed in June 2002. The reform is expected to have
benefited all natives regardless of their place of residence in Switzerland similarly. Specifically, it is unlikely
to have promoted the commuting of Swiss nationals from border areas due to the relatively high living costs
combined with high wage levels and low unemployment in Switzerland.

7
 worker (local priority requirement).
The AFMP was implemented in three steps. From June 2002 onward, cross-border
commuters from EU-15 and EFTA countries were free to reside outside the border zones
of the home country. In addition, they were required to return to their place of residence
only once a week rather than every day. The work permit was no longer bound to a specific
job, and its validity was extended to the length of the working contract, for a maximum
of five years. In June 2004 the local priority requirement was abolished, and as a result,
cross-border commuters could be hired under the same conditions as resident workers in the
Swiss border zones. Full liberalization across the entire country came into force in June 2007
Downloaded from by guest on September 4, 2025. Copyright 2023

when commuters were allowed to work anywhere in Switzerland. Interim regulations applied
for other EU member states and were relaxed over time.
The number of cross-border commuters increased sharply after the free movement reform
was implemented. As administrative data show, most of the commuters work in commuting
zones in the border area, where their share in total employment rose from 9.9% in 2001
to 14.2% in 2017. In the latter year, 95% of all cross-border commuters were nationals
of neighboring countries Austria, France, Germany, or Italy. Consistent with travel costs
depending on distance, commuters generally work in commuting zones close to their place
of residence where the same language is spoken.2
Table 1 shows how native workers differ from cross-border commuters. The average age of
both types of workers is around 40 years, while women are underrepresented in the group of
commuters. Native workers are less likely to have at most a lower-secondary education than
commuters (11.1% versus 26.3%) and are more likely to have an upper-secondary education
(64.3% versus 49.5%). The share of the tertiary educated is similar among the two types of
workers. The likelihood of holding a position with managerial responsibilities is higher among
native workers, while commuters are overrepresented among STEM workers. In Section 4
2
Around 97% of Austrian and 98% of German commuters work in a Swiss municipality in which German
is spoken by most residents. The share of Italian and French commuters who work in Italian- and French-
speaking Swiss municipalities is 88% and 80%, respectively.

8
 we explore how these shares evolve in affected and non-affected areas after the AFMP was
implemented.

2.2 Dual Education System

The Swiss education system has features common to other European countries. At the upper-
secondary level, one can follow two types of tracks: a vocational or specialized education
track and a general education track (see Online Appendix Figure A1). According to the Swiss
Federal Statistical Office (FSO), 66.6% of upper-secondary students pursue a vocational
Downloaded from by guest on September 4, 2025. Copyright 2023

education in 2019, 6.8% a specialized education, and 26.6% a general education (Bundesamt
für Statistik, 2019a). In the vocational track, an individual works three to four days a week as
an apprentice and attends school the rest of the time. In the specialized track, an individual
attends general and vocational school classes and does an internship. In the analysis, we refer
to both tracks as vocational because they prepare students to enter the labor market or enroll
in tertiary education. In contrast, the objective of the general track is to prepare students
for enrollment in tertiary education only. To enter tertiary education, a student must pass
a matura examination at the end of upper-secondary education. As of 2019, 22.1% of Swiss
residents younger than 25 years hold a general matura, 15.9% a vocational matura, and
3.4% a specialized matura (Bundesamt für Statistik, 2019b). Sixty-four percent of the 2012
upper-secondary graduates with a vocational matura and 84% of those with a specialized
matura enroll in tertiary education within 42 months after graduating. This is significantly
lower compared to 94% of those with a general matura (Bundesamt für Statistik, 2018).
Two types of tertiary institutions exist: universities and federal institutes of technology
(referred to as universities or UNIs) and universities of applied sciences (UAS). UNIs are the
oldest institutions with a right to grant tertiary-level degrees, and the UAS were established
in 1997.3 Unlike UNIs that are committed to a combination of teaching and research, UAS
impart professional skills with a focus on practice and application. Roughly 60% of all
3
UAS are organized by cantons. In some cantons the UAS include teacher education, while other cantons
have set up independent teacher education universities. We combine these institutions with the UAS.

9
 students are enrolled at UNIs in 2017 and 93.9% of them have a general background at
the upper-secondary level. At UAS, 80.2% of all students have a vocational education.
Both types of institutions offer STEM and non-STEM education. Around 58.1% of all UNI
students are enrolled in a non-STEM field in 2017; this share is close to 63.3% at UAS.
The Swiss education system offers a unique setting to investigate enrollment decisions
as the lack of supply constraints enables us to infer demand for tertiary education from
enrollment. Besides a matura, no major entry restrictions exist for Swiss nationals at the
undergraduate level. A general matura typically grants access to any degree in the chosen
university. As an exception, health degrees can have a cap on the number of students
Downloaded from by guest on September 4, 2025. Copyright 2023

enrolled in a year. UAS can require a certain major of the vocational matura or relevant
work experience to enroll in specific fields, and they often conduct interviews to test the
ability of candidates in social- or health-related fields. There is overall little screening at
entry, but the pool of eligible students is already selected due to the admission requirements
for upper-secondary education tracks resulting in a matura.
Figure 1 shows the locations of the tertiary education institutions across Switzerland
in 2017. Most of the institutions are in the northern and western part of the country and
are clustered in the urban centers. The 12 UNI institutions are spread over ten cities. In
contrast, most of the UAS have several locations, which are often specific to a study field.
The high density of institutions we observe in 2017 enables a large share of the population
to commute daily to classes, lowering the costs of studying.4 UAS expanded considerably
across the country during our study period. We will control for this change in the education
supply in our analysis.
4
Yearly study costs are estimated to be around CHF 24,000, including tuition fees that are generally
below CHF 2,000 for Swiss nationals. See, for example, the calculations by the study advisory service from
the University of Zurich. On January 5, 2023, one Swiss franc (CHF) was equivalent to approximately USD
1.1.

10
 3 Data and Methods

3.1 Data

In the enrollment analysis, we use administrative data from the Swiss Higher Education
Information System (SHIS-studex). This is an individual-level database covering all matric-
ulated students at UNIs and UAS in Switzerland. It includes students at UNIs since 1990
and UAS since their establishment in 1997. The variables used are age, gender, nationality,
type of certificate granting access to tertiary education (matura), place of residence at the
time of taking the matura (pre-enrollment place of residence), type of tertiary institution,
Downloaded from by guest on September 4, 2025. Copyright 2023

and field of study. The structure of the SHIS-studex dataset allows us to track individuals
from the point of enrollment up to graduation and provides information on received degrees.
More information is available in the Online Appendix 3.
We are interested in demand for tertiary education and focus on first-year students en-
rolled in a bachelor’s program over the period 1997–2017. We do not include immigrant
students in our baseline sample because their enrollment decisions may depend on different
factors than the decisions of Swiss nationals (for example, Schündeln, 2014). We assign the
native tertiary students to the commuting zone of residence at the time of receiving their
Nr first-year studentsct
matura. We calculate our main outcome as Enrollment ratect = Birth cohort sizect
, where c
is the commuting zone. The cohort is the Swiss population at the median age of first-year
students, which is 21. To measure the cohort, we use information about the size of the
native population at the municipality level and the age structure of the population at the
cantonal level from the FSO. We add to the SHIS-studex dataset the geographic location of
the tertiary institutions, which we collected from the institutions’ websites.
Additionally, we use information from the Survey of Higher Education Graduates (EHA),
which is conducted every two years. The survey has a panel structure where individuals
respond to questions related to their working experience and acquired skills one and five years
after graduation. We use first-wave responses because we are interested in outcomes shortly

11
 after graduation. We consider the subset of Swiss graduates with a bachelor’s or master’s
degree who also have a Swiss matura. We then use information about the commuting zone
of residence (current and at the time of taking the entrance exam), commuting zone of work,
and the mapping between fields of study and occupations.
In the labor market analysis, we rely on the Swiss Earnings Structure Survey (SESS),
which is a large-scale firm survey conducted every two years. It is a repeated cross-section of
private sector firms in the secondary and tertiary sectors of the economy. Our sample covers
the years 1996–2016. We use information on the firm location at the commuting zone level,
which is the most detailed geographical unit available. We limit the sample in the main
Downloaded from by guest on September 4, 2025. Copyright 2023

specification to employees aged 18–40 and in the robustness checks to employees aged 18–65.
To calculate the share of cross-border commuters, we divide the number of commuters by the
total number of employees. To calculate our labor market outcomes, we use data on native
gross hourly wages, the level of managerial tasks, and the number of native employees. We
differentiate three types of education based on the highest level attained — up to lower-
secondary, upper-secondary, and tertiary training. We distinguish positions with medium
or senior management responsibilities from positions with no management, and junior or
executive management responsibilities. Similarly, we differentiate between workers in STEM
and non-STEM occupations. When looking at occupations, we limit the observation period
to 1996–2010 because different occupation classifications were used before and after 2010.
We also use administrative data from the Cross-Border Commuters Statistics and publicly
available employment data at the municipality level for 1995, both provided by the FSO.
Additionally, we collected travel time data for each municipality from www.map.search
.ch, which we accessed in December 2018. We take the travel time by car from each mu-
nicipality m to the closest border crossing or a border checkpoint according to the Fed-
P
eral Customs Office and calculate the measure Travel timec = m∈c Travel timem,2018 ×
Nr employedm,1995
Nr employedc,1995
. Commuting zones with a border crossing or border checkpoint are assigned
a value of zero minutes. The Federal Office of Topography (swisstopo) provides the com-

12
 muting zone shapefile that we use.

3.2 Empirical Strategy

Motivated by the nature of the policy change, the empirical analysis is based on a standard
difference-in-differences strategy. We investigate the reform effects by comparing commuting
zones close to the border with those further away before and after the regulatory change.
Figure 2 shows that exposure to commuters declines sharply with travel time from the
border. In the main part of the analysis, we use a fixed threshold to define treatment, which
Downloaded from by guest on September 4, 2025. Copyright 2023

is consistent with Beerli et al. (2021). Specifically, we consider commuting zones closer than
a 30-minute commute as affected by the reform and those further away as not affected.
This approach assigns 35 out of the 106 commuting zones to the affected group and the
remaining 71 to the non-affected group (see Figure 1). There is no discontinuity in exposure
to cross-border commuting at the 30-minute threshold. To take this into account, we consider
different treatment assignments in alternative specifications. In one of these specifications,
we use a continuous treatment definition based on distance to the border according to the
formula exp(−0.05 × travel time). Figure 2 shows that the function follows closely how
observed commuting flows vary with distance.5
We run the following specification in the main part of the analysis:

yct = α+β1 Transitiont × 1(Travel timec ≤ 30 min)+

(1)
β2 Postt × 1(Travel timec ≤ 30 min) + X0ct γ + δc + δnt + εct ,

where c is a commuting zone, n is a NUTS 2 region, and t is a year. In the enrollment

analysis, c refers to the commuting zone of residence at the time of receiving the matura. In
the labor market analysis, c refers to the commuting zone of the workplace.
5
Figure 2 also reveals that commuters work further away from the border in 2017 than they did in
1997. The continuous function tracks well the relationship between commuter flows and travel time in both
years and therefore accounts for the upward trend in the commuting distance. Criticism of difference-in-
differences regressions with a continuous treatment has been recently raised by Callaway et al. (2021) as
stronger assumptions tend to be required compared to the binary treatment case.

13
 Our main outcome is the share of native first-year students in the birth cohort. We
also look at the native gross hourly wage rate, the probability of holding a managerial
position, and the number of those employed. To ensure that any differences in wages and
the probability of holding a managerial position are not driven by compositional changes,
we use residual values. These are obtained by regressing individual-level gross wages and
an indicator for holding a middle or upper management position, respectively, on age, age
squared, gender, and educational attainment controls. We then calculate the average value
of the residuals within a commuting zone for each survey year (Edo, 2019; Peri and Yasenov,
2019).
Downloaded from by guest on September 4, 2025. Copyright 2023

We estimate the reform effect by distinguishing between three periods: pre-reform (1997–
2001), transition (2002–2006), and post-reform (2007–2017). The observation period for the
labor market outcomes is 1996–2016 due to data availability with the pre-reform period
between 1996 and 2000 and the post-reform period between 2008 and 2016. The coefficients
of interest, β1 and β2 , show the difference in the dependent variable between affected and
non-affected commuting zones during and after the reform compared to pre-reform years.
We include commuting zone fixed effects to capture local variation in the outcomes of
interest. The Swiss education system is organized at the cantonal level, with the catchment
area of a tertiary institution typically extending over several cantons.6 Changes in an in-
stitution’s policy are therefore likely to affect geographic clusters of commuting zones in a
similar way. To take this into account, we include NUTS 2 region × year fixed effects δnt . In
the baseline enrollment analysis, the vector of control variables X includes the natural log of
native population. Additional variables that could vary during the period and across com-
muting zones are introduced in robustness checks. Consistently, in the labor market analysis
we also add commuting zone and NUTS 2 region × year fixed effects. When estimating
Equation 1, we use weights fixed in the first year of the analysis to account for the different
6
Switzerland has seven NUTS 2 regions, and three commuting zones fall into more than one NUTS 2
region. We allocate the commuting zones to the NUTS 2 region in which the majority of the commuting
zone’s population resides. The share of enrolled undergraduate students who come from the same NUTS 2
region as the one where the institution is located is, on average, 60.7% in 2017.

14
 population and employment sizes across commuting zones. Standard errors are clustered at
the commuting zone level.
Graphically we show yearly estimates βt from the following event study:

2017
X
(2) yct = α+ βt Y eart × 1(Travel timec ≤ 30 min) + X0ct γ + δc + δnt + εct .
t=1997

The event study approach allows us to investigate the timing of the enrollment and labor
market responses. The coefficients βt capture the impact of the reform relative to the last
year in the pre-reform period. This is 2001 in the enrollment analysis where we have yearly
Downloaded from by guest on September 4, 2025. Copyright 2023

data. In the analyses on commuter exposure and labor market conditions with biennial data,
the reference year is 2000.
The key assumption under which our results are valid is that enrollment rates and labor
market conditions would have followed the same trend in affected and non-affected areas
absent the reform. We compare yearly coefficients in the pre-reform period to investigate
whether this assumption is likely to hold and find no evidence to the contrary. Pre-trends
for overall and UNI enrollment are shown for 1991–2001 and for UAS enrollment, they cover
the period since their establishment in 1997. Similarly, the results are robust to including
additional control variables that could have evolved differently over time in the two groups
of commuting zones. These results are discussed in more detail in Section 4.
The common trends assumption could be violated if other policy changes occurred during
the study period and affected individuals residing in the affected and non-affected commut-
ing zones differently. In the 2000s the Bologna reform was introduced, which promoted
the international mobility of higher education students. The tertiary education systems in
Switzerland and its neighboring countries were already similarly structured before the re-
form. This makes it unlikely that natives from commuting zones closer to the border more

15
 often enroll into tertiary institutions in border areas abroad due to the Bologna reform.7 If
anything, we expect that the reform increased enrollment into foreign institutions for natives
coming from affected commuting zones, which would bias our estimates downwards.
The Stable Unit Treatment Value Assumption (SUTVA) is the second important iden-
tifying assumption. We are interested in local labor market conditions and their impact on
demand for education and take the commuting zone as the unit of observation. A com-
muting zone has, on average, approximately 80, 000 residents and 49, 000 workers in year
2017. Commuting zones are considered small-scale labor markets and are constructed by the
FSO as clusters of municipalities where a large share of the population resides and works.
Downloaded from by guest on September 4, 2025. Copyright 2023

Specifically, the share of employed who reside in the commuting zone where they work,
on average, 64.2%, and shares in the group of affected and non-affected commuting zones
are virtually the same. Overall, the commuting zone reduces concerns about geographical
spillovers because local labor market conditions are most relevant to individuals who reside
in that region. Our approach is consistent with the literature showing that job search is
local and the attractiveness of jobs declines sharply with distance (Manning and Petrongolo,
2017).
We argue that the labor market conditions in the pre-enrollment place of residence are
major determinants of demand for education. This is consistent with Long et al. (2015),
who show that local changes in earnings have a stronger impact on major choices in higher
education than changes at the national level. Acton (2021) argues that changes in local
employment conditions shape the labor market expectations of individuals enrolling into
community college programs that are closely tied to specific occupations similar to UAS
degrees. In our setting, own labor market experience as gained during the vocational ed-
ucation is likely local due to limited mobility at that age. As students tend to enroll into
tertiary education shortly after graduating from upper-secondary education, the labor mar-
7
The following are tertiary institutions located in the neighboring countries close to the Swiss border: the
University of Konstanz and the Zeppelin University in Germany; universities of applied sciences in Dornbirn
and Feldkirch in Austria; the University of Liechtenstein.

16
 ket experience accrued during the vocational training is likely decisive. Another important
information source for educational decisions is the experience of immediate family members
(Xia, 2016), which is also arguably accrued locally.
The EHA survey allows us to investigate location choices after graduation relative to
the pre-enrollment place of residence. In 2017, 59.0% of the graduates live in the same
commuting zone where they resided during their upper-secondary education. Shares for the
affected and non-affected areas are essentially the same. Almost 60% of all graduates report
that finding a job in the local area is important or very important to them. The share of
respondents who work in the same commuting zone where they resided before enrollment is
Downloaded from by guest on September 4, 2025. Copyright 2023

29.4%. The share is slightly higher in affected than in non-affected commuting zones, which
does not point to an increased avoidance of foreign competition as a result of the reform.
Because acquiring tertiary education is shown to increase geographical mobility (Malamud
and Wozniak, 2012), these numbers should be conservative proxies for the ex-ante intention
to return to the pre-enrollment place of residence. If, at the time of enrollment, individuals
consider labor market outcomes other than those in the commuting zones where they reside,
our estimates would tend towards zero by reducing the difference between those coming from
affected and non-affected areas.

4 Results

In this section, we first investigate the reform effect on the inflow of cross-border commuters.
Then we show our findings on native enrollment by institutional type and enrollment by field
of study.

4.1 Exposure to Cross-Border Commuters

To justify the treatment assignment rule, we estimate Equation 1 and compare the share of
cross-border commuters in total employment across affected and non-affected areas in the

17
 different periods. Column 1 of Table 2 shows that commuting zones within 30 minutes of
travel time from the national border experienced a large inflow of commuters relative to
those further away after the AFMP was introduced. While average exposure in the affected
commuting zones grew from 14.4% in the pre-reform period to 18.6% in the post-reform
period, we estimate a reform effect of 3.3 percentage points after controlling for commuting
zone fixed effects and NUTS 2 × year fixed effects. Magnitudes increase during the post-
reform period as shown in Figure 3a. The continuous rise in exposure to commuters highlights
the permanent nature of the reform. Online Appendix Figure A2a replicates these results
with administrative data; estimates are larger in magnitude as we fix the denominator in
Downloaded from by guest on September 4, 2025. Copyright 2023

1995 for which employment data are available. The results show that cross-border commuting
was already slightly on the rise in the last years of the pre-reform period. This could be
explained by an informal relaxation of regulations before 2002, which we take into account
when discussing the timing of the enrollment results.
In Columns 2–4 of Table 2, and in Figures 3b–3d, we look at exposure to cross-border
commuting by educational level. We find that the rise in the share of cross-border com-
muters among the upper-secondary educated is 4.6 percentage points. Among the tertiary
educated, the rise is 3.2 percentage points in the post-reform period. The positive effect
on the former group is already significant during the transition period. We do not find a
significant increase in commuting of lower-secondary-educated workers as presented in the
table, while the positive estimates in Figure 3b are driven by the choice of the reference year.
In Online Appendix Table A1 we test the sensitivity of the results to lower and higher
threshold values in the treatment definition. We find that the estimated magnitude of the
commuter inflow declines as we allocate commuting zones further from the border to the
affected group. As a generalization, we confirm the rise in cross-border commuting using the
continuous treatment measure. Another concern we address is whether resident immigrants
are, like commuters, more often employed in commuting zones close to the border in the
post-reform period. Online Appendix Figure A2b shows that the share of immigrants does

18
 not evolve differently across affected and non-affected areas during the study period. We
therefore focus on cross-border commuters as the relevant group of foreign workers given our
empirical strategy.
Next, focusing on skilled individuals, we study whether cross-border commuters are over-
represented among workers with certain types of skills. We start by linking fields to occu-
pations and create the variable Sh employedj , which proxies the share of employees trained
in field j:
Downloaded from by guest on September 4, 2025. Copyright 2023

O
X
(3) Sh employedj = Sh employedo × Sh employedoj , j ∈ [1, 22] ,
o=1

where Sh employedoj is the share of employed individuals in occupation o with a degree in

field j, which we multiply by the share of employed in the same occupation Sh employedo .
Intuitively, we allocate individuals employed in an occupation to fields of study and take
into account the occupation’s size.
We infer the link between study fields and occupations from their joint distribution
provided by the EHA survey (2003–2017). This approach is consistent with the fact that
natives do not observe the education of commuters but have knowledge of their occupations.
We add the distribution of cross-border commuters and residents across occupations from
1999 and 2000 administrative data, respectively. These years are the earliest available and
hence alleviate concerns about endogenous adjustments in the native occupational choices
to changes in the skill levels of commuters or vice versa. Finally, we divide the skill supply
of cross-border commuters by the skill supply of resident workers in a field:

Sh cross-border commutersj
(4) Relative skill supplyj = , j ∈ [1, 22] .
Sh residents employedj

The Relative skill supplyj measure shows how the high-skilled commuters are allocated across

19
 study fields j relative to the high-skilled workers residing in the country. A higher value of
the measure implies that commuters are relatively more likely to have received training in
this specific field than resident workers.
In Column 2 of Table 3, we present the skill supply of commuters relative to that of
resident workers for each study field. The least affected fields – those with the lowest ratio
– are listed first, and the most affected fields come last. Frontier workers are more often
trained in architecture and construction, ICT, and forestry, which are study fields that build
technical and numerical skills. Commuters are underrepresented in education, languages,
and law, which are fields that build knowledge less transferable across borders and require
Downloaded from by guest on September 4, 2025. Copyright 2023

social skills or a high level of language skills. If we divide the study fields based on the
variable Relative skill supply into affected (value above one) and non-affected (below one),
we see that the former group coincides with STEM and the latter with non-STEM fields.
There are two exceptions: the arts is a non-STEM subject classified as affected, and health
is a STEM field classified as non-affected.
In Table 4, we study the variation in the commuters’ skills over time and complement
the static picture of the skill distribution presented in Table 3. Specifically, we investigate
the change in exposure to cross-border commuters by both education and occupation. We
consider upper-secondary and tertiary levels of education, and we split occupations into
STEM and non-STEM. At both levels we observe a stronger inflow in STEM than in non-
STEM occupations in the transition and post-reform periods. The literature has established
a similar specialization pattern between native and immigrant workers (Hunt and Gauthier-
Loiselle, 2010; Hanson and Slaughter, 2018). We confirm this finding in a context where
foreign workers are culturally and linguistically similar to the natives, as discussed in Section
2.
To summarize, our findings on the effect of the AFMP on cross-border commuting are
consistent with Beerli et al. (2021). In addition, they show evidence that the reform had a
positive effect on labor demand in affected areas because additional jobs were created for the

20
 new foreign workers. Productivity and innovation rise mainly at firms that were previously
unable to recruit skilled workers. The greater availability of cross-border workers also en-
couraged capital formation as proxied by firm entry. These changes come from skill-intensive
sectors such as high-tech manufacturing and knowledge-intensive business services. As these
industries employ skilled individuals, changes on the demand side can affect labor market
conditions relevant to natives who are about to decide whether to enroll in tertiary educa-
tion or not. Ultimately, both supply and demand shifts likely affect how native conditions
respond to the policy change. Our estimates are thus interpreted as reform effects.
Downloaded from by guest on September 4, 2025. Copyright 2023

4.2 Enrollment by Institutional Type

Summary statistics in Table 5 show that during our study period, average enrollment in
tertiary education is higher in commuting zones affected by the reform’s introduction than
in those not affected. This difference comes from enrollment at UNIs. Enrollment rates are,
on average, similar at UAS in the two groups of commuting zones, but Online Appendix
Figure A3 shows that enrollment grew faster in affected commuting zones after the reform
was implemented. We next test whether these patterns are statistically significant and persist
conditional on fixed effects and control variables.
The estimates in Column 1 of Table 6 show a positive but insignificant rise in overall
enrollment in the post-reform period among individuals residing in affected commuting zones
before beginning their studies compared to non-affected commuting zones. The responses
differ by institutional type. Results in Column 3 indicate that individuals from affected areas
enroll significantly more often at UAS. The magnitude of the effect is 1.1 percentage points.
Average enrollment rates in the affected areas increase from 7.8% in the pre-reform period
to 18.4% in the post-reform period. The reform effect can account for almost 10.4% of the
enrollment growth observed during the period and is 14.1% of the pre-reform enrollment
level. In contrast, in Column 2 we find no change in entry into UNIs between the affected
and non-affected areas.

21
 Figure 4 shows that demand for tertiary education, overall and by institutional type,
evolves similarly between the affected and non-affected commuting zones in the pre-reform
years. This suggests that the common trends assumption is likely to hold. Indeed, the timing
of the increase in enrollment at UAS is in line with the intensity of the inflow of cross-border
commuters presented in Figure 3a. Although we observe a small increase in commuting in
the pre-reform period, we find that enrollment goes up only in the post-reform period, when
all barriers were abolished, and the inflow of frontier workers was significant.
Panels A and B of Online Appendix Table A2 show that the threshold of 30 travel
minutes is not decisive for the main findings. Moreover, the estimates remain similar when
Downloaded from by guest on September 4, 2025. Copyright 2023

using the continuous travel time measure (Panel C). Online Appendix Table A3 investigates
whether the rise in UAS enrollment is sensitive to different control variables and the weighting
scheme. Changes in the supply of education and demand for labor could be confounding
factors to the common trend assumption. We test whether the rise in enrollment rates is
driven by the availability of new study locations and study fields because our observation
period coincides with the expansion of UAS.8 Column 2 shows that the result is robust to
controlling for the presence of UAS as well as the number of study fields offered at UAS
within a radius of 20 kilometers.9 Note that the reform estimate is of similar size to the
effect of a change from no to at least one UAS institution within this radius. There may be
concerns that labor demand evolved differently between affected and non-affected commuting
zones due to factors such as export demand (Atkin, 2016). We therefore include a Bartik-
type measure of employment, relying on the industrial composition of each commuting zone
in 1995, and aggregate annual employment growth at the industry level (see Bartik, 1991).10
8
Hoxby (2009) finds, for the United States, that university choice is less driven by distance in recent times
partly due to declining transportation costs. In the context of Switzerland, Denzler and Wolter (2010) argue
that the distance to university matters for both the decision to enroll and the study field choice, particularly
for individuals from middle and low socio-economic groups.
9
We calculate the 20 kilometers as the distance from the largest municipality in a commuting zone in
1990.
10
We construct the Bartik variable as follows: Bartikct = i Sh employedic1995 × NNr Employed
r Employedit
P
i1995
, where
i denotes the industry, c denotes the commuting zone, and t denotes the year. The industry definition follows
the General Classification of Economic Activities (NOGA-08) at the two-digit level.

22
 As reported in Column 3, controlling for labor demand does not change the results compared
to our baseline specification. Additionally, in Columns 4 and 5 we confirm that the control
variables, the NUTS 2 region × year fixed effects, and the weights do not drive the results.
In Online Appendix Table A4 we redefine our outcome variable. We investigate the
reform effect on the natural log of the number of natives enrolled and confirm that the effect
is not driven by the variation in the cohort’s size over time in Column 2. In Column 3
we look at the enrollment rate of native and immigrant students and show that results are
consistent with our baseline measure. In the last column, we test if there is an impact on
the graduation rates of the first-year students. We compute the graduation rate of students
Downloaded from by guest on September 4, 2025. Copyright 2023

Nr graduates by 2017ct
as Graduation ratect = Nr first-year studentsct
, where t is the first year of enrollment. We focus
on first-year students who enrolled in the period 1997–2013, which leaves them with at least
four years to complete a three-year bachelor’s degree. We find no evidence for significant
differences in graduation rates between affected and non-affected areas among those who
enrolled shortly after the AFMP was implemented.
We conduct a heterogeneity analysis to investigate selection. We examine if natives
respond differently by pre-determined variables such as age, gender, and characteristics of
their pre-enrollment place of residence. The results in Online Appendix Table A5 for UAS
show that the rise in enrollment in the post-reform period is driven by individuals who are
of relatively young age (21 and younger). This finding is in line with Wiswall and Zafar
(2015), who show that younger students are more responsive to changes in future earnings
than older students. The opportunity cost of studying in terms of forgone wages is arguably
lower for the younger group. The estimates are statistically significant for male and not for
female enrollment, but the difference between the two coefficients is not significant. The
same holds for those coming from urban and rural commuting zones, and from different
language regions.

23
 4.3 Enrollment by Field of Study

Cross-border commuters are more likely employed in STEM occupations than natives as
shown in Tables 3 and 4. Online Appendix Figures A3d and A3e plot raw enrollment rates
into STEM and non-STEM fields at UAS and show that demand for non-STEM fields grew
faster in affected relative to non-affected commuting zones in the post-reform period. We
next test whether these changes are statistically significant using the difference-in-differences
specification.
The estimates in Panel C of Table 7 and Figure 5 confirm the rise in enrollment in non-
Downloaded from by guest on September 4, 2025. Copyright 2023

STEM fields. We estimate a 0.8 percentage points increase in the post-reform period that
can account for 9.8% of the increase in enrollment observed during the period and is almost
21% of the pre-reform average. The evidence from the analysis of enrollment in non-affected
fields is consistent. Ransom and Winters (2021) estimate that natives are crowded out of
STEM fields in regions with more foreign workers, but we find no such evidence. When we
split the group of non-STEM fields into separate ones following the International Standard
Classification of Education (ISCED) classification, we find that the positive post-reform
effect at UAS comes from welfare, business, and arts (see Online Appendix Table A6). We
argue that the skills especially acquired from a business study could be complementary to
the STEM skills brought by the commuters. Panels A and B of Table 7 show no evidence of
a change in enrollment by field of study overall and at UNIs, respectively.
In Online Appendix Table A7 we show that the overall increase in the demand for non-
STEM and non-affected fields is robust to variations in the threshold that defines the affected
area. Enrollment in STEM and affected fields turn significant at lower and higher thresh-
olds than 30 minutes and in the continuous specification. Overall, the rise in non-STEM
enrollment is robust and of larger magnitude compared to the STEM enrollment results that
depend on the treatment definition. Online Appendix Table A8 reports results from specifi-
cations including additional control variables in Columns 2–3, without control variables and
NUTS 2 region × year fixed effects in Column 4, and without weighting scheme in Column

24
 5. The estimates remain very close to the ones from the main specification. In an unreported
analysis we confirm that the results are also robust to alternative definitions of the outcome
variable analogous to Online Appendix Table A3 for UAS enrollment.

5 Mechanisms

Natives respond to the free movement reform by acquiring more tertiary education at UAS.
When faced with stronger foreign competition in the labor market, education offers an op-
portunity to stay competitive and to benefit from complementarities by specializing in fields
Downloaded from by guest on September 4, 2025. Copyright 2023

that foreign workers are less likely to be trained in. In this section, we explore mechanisms
related to labor market outcomes by building on the work of Beerli et al. (2021). We examine
native wages, the probability of holding a managerial position, and employment by educa-
tion level and occupation and discuss the timing of these effects relative to the timing of the
enrollment responses. We distinguish first-year students by their educational background
at the upper-secondary level to test whether previous labor market experience affects how
responsive enrollment decisions are to changing labor market conditions. Below, we present
evidence that enrollment decisions are consistent with changes in native labor market out-
comes. Then, we consider an alternative mechanism that relates to the enrollment of foreign
students, but we do not find evidence that it plays a role in our setting.

5.1 Native Labor Market Conditions

The inflow of commuters into local labor markets is likely to affect natives’ labor market
conditions. We focus on young native workers as changes in their labor market conditions
are likely most relevant for the educational decisions of upper-secondary graduates. Panel A
of Table 8 reports a decrease in wages for upper-secondary-educated workers that is driven
by the late post-reform period between 2012 and 2016. In contrast, we find evidence of an
increase in wages for tertiary-educated workers, which is largest in the early post-reform

25
 period between 2008 and 2010.11 To better understand responses at the study field level,
we examine native wages by education and occupation. Columns 1 and 2 in Panel A of
Table 9 show that at the upper-secondary level in the post-reform period the returns tend
to rise in STEM occupations and fall in non-STEM occupations. The positive estimate for
workers with a tertiary education is driven by non-STEM occupations, though the results are
insignificant with a p-value of 0.11. This evidence is consistent with the rising demand for
non-STEM tertiary education and with the literature showing that future earnings matter
for major choice (Schweri and Hartog, 2017; Abramitzky et al., forthcoming).
Wage effects could be hard to observe, especially for young individuals. In addition,
Downloaded from by guest on September 4, 2025. Copyright 2023

the study field choice elasticity has been found to be often relatively low (Patnaik et al.,
2021). Changes in the type of positions that natives hold are important indicators for
career prospects and may be more visible. In Panel B of Table 8 we look at the probability
that natives hold at least middle management positions and find that the share of tertiary-
educated natives holding such positions tends to increase starting in the transition period
in affected commuting zones, but these results are statistically significant only in the full
sample of workers (see Table A9). We find that the increase is driven by those employed in
non-STEM occupations (Panel B of Table 9). This pattern is consistent with the changes in
wages for tertiary-educated workers.
Changes in the number of natives employed with certain types of skills can be another
visible signal of improving or worsening labor market conditions. Our results in Panel C
of Tables 8 and 9 show that the number of native workers by education, and by education
and occupation, do not evolve differently during the observation period in affected and non-
affected areas.12 Consistently, Online Appendix Figure A4 shows no differential trend in
native population size between affected and non-affected areas. Therefore, native reloca-
11
Note that the estimates based on all employees are provided in the Online Appendix (see Tables A9 and
A10) and are consistent with the results based on the subset of young workers.
12
In an unreported analysis we also look at the native employment and unemployment rates by education
based on data from the Swiss Labor Force Survey from the FSO. We do not find that these outcomes evolved
differently between affected and non-affected areas during our observation period.

26
 tion away from the border areas due to the reform should not pose an issue to the effect
identification (Borjas, 2006).
We observe a higher demand for tertiary education at UAS. In 2017, around 80% of
first-year students at these institutions have a vocational background. We show in Table 10
that the rise in enrollment at UAS is driven by students with a vocational background.13
The response of the vocationally educated can be explained by their higher awareness of
changes in local labor market conditions compared to the generally educated. Labor mar-
ket conditions of upper-secondary-educated workers are directly relevant to natives with a
vocational background because of their outside options. Moreover, a vocational training
Downloaded from by guest on September 4, 2025. Copyright 2023

likely increases the level of knowledge about returns to tertiary education due to its focus on
practical experience in contrast to general education. We also investigate the type of study
field at the upper-secondary level and find in Column 4 of Table 10 that individuals with a
non-STEM background react to the reform. We conclude that access to tertiary education
enables natives to upgrade their non-STEM skills and benefit from the rising wage premium
for non-STEM tertiary degrees.
To summarize, our labor market analysis suggests that the increase in non-STEM enroll-
ment at UAS can be attributed to lower opportunity costs of studying, higher returns to a
tertiary degree, or to a combination of the two factors. The decrease in wages for upper-
secondary educated workers coincides with the period when native enrollment increased the
most. The increase in wages for tertiary educated workers started before the rise in native
enrollment and persisted after the reform was fully implemented. Overall, timing suggests
that the rise in native demand for tertiary education is a response to recent shocks in the la-
bor market. Thus, higher education enables natives to take advantage of complementarities
between native and foreign workers.
13
This link to the professional world persists during the studies: students at UAS report more often that
they work while studying (79%) compared to those at UNIs (69%) (Bundesamt für Statistik, 2020).

27
 5.2 Foreign Student Enrollment

We next test an alternative mechanism that relates to the role of foreign student enroll-
ment. The literature on university enrollment and study field choice has investigated the
link between the presence of foreign students and natives’ decisions. Recent studies find, on
average, no or a positive effect on native enrollment (Machin and Murphy, 2017; Shih, 2017).
Earlier studies also document crowding-out effects (Borjas, 2004). At the field level, there
is some evidence that foreign students reduce the likelihood that natives major in a STEM
field (Orrenius and Zavodny, 2015; Anelli et al., forthcoming). Unlike in the US and the
Downloaded from by guest on September 4, 2025. Copyright 2023

UK, where these studies were conducted, Switzerland has generally no cap on the maximum
number of students enrolled that could explain crowding out. Moreover, tuition is, to a large
degree, publicly funded, which makes it unlikely that there is cross-subsidization of native
students through tuition fees paid by foreign students. We conclude that the mechanisms
discussed in these studies are not directly applicable in the Swiss context.
Our empirical strategy further mitigates concerns that natives’ educational decisions are
affected by foreign students. We measure the overall demand for two types of institutions
and for broad groups of study fields. This alleviates the potential crowding-out or crowding-
in effect at the institution × narrow field level as switching between individual institutions
and related fields can help students to avoid or to find more foreign peers. At the same
time, the share of foreign students in all students enrolled at the bachelor level is sizeable
in Switzerland (close to 19% in 2017). In the rest of this subsection, we consider the role of
immigrant students — non-Swiss with a Swiss entry exam — and international students —
non-Swiss without a Swiss entry exam to tertiary education.
Immigrant students have contact with natives in their pre-enrollment place of residence
and also within the educational system before enrolling into tertiary education. They rep-
resent about 7% of total enrollment in 2017, and this number is similar at UNIs and UAS
and in STEM and non-STEM fields. If their enrollment evolved differently in affected and
non-affected areas during the study period, this could have an effect on native demand for

28
 education. We first test if the reform had a direct effect on immigrant student enrollment.
In Online Appendix Table A11 we look at enrollment among first-year immigrant students
and observe a drop during the transition period driven by UNIs. This result should be taken
with caution given that the negative trend started during the pre-reform period. Moreover,
immigrants are a relatively small group of individuals who migrate for a variety of reasons
and are likely to selectively move into areas of the country. Given these results, we do not
expect that the reform affects native enrollment through a change in the student composi-
tion at UAS. Second, we test whether our main result is robust to controlling for changes in
immigrant enrollment by institutional type or field. by changing immigrant enrollment by
Downloaded from by guest on September 4, 2025. Copyright 2023

institutional type or field. The estimates in the last column of Online Appendix Tables A3
and A8 show that higher immigrant enrollment at UAS and in non-STEM at UAS, respec-
tively, is correlated with higher native demand for education, but our coefficient of interest
remains very close to the baseline specification.
The share of international students in total enrollment was approximately 12% in 2017.
In contrast to immigrant students, international students arrive in Switzerland at the time
when they enter tertiary education. They reside abroad when taking their entry exam to
tertiary education, so it is not possible to allocate them to a commuting zone of residence
and replicate the analyses performed with immigrant students. As natives are not exposed
to international students in lower education levels, it is unlikely that they are a factor in their
enrollment decisions. Most importantly, for international students to present a challenge to
our empirical strategy, they must impact natives coming from the affected and non-affected
areas differently. In addition, this impact needs to vary over time in line with the timing of
the free movement reform. As the share of international students at UAS stayed around 9%
between the early 2000s and 2017, we consider it unlikely that their presence plays a role in
explaining our results.

29
 6 Conclusion

We examine the impact of the introduction of free movement of workers on native demand
for tertiary education in Switzerland. Our findings reveal that individuals from commuting
zones with high exposure to cross-border commuters enroll more often at universities with
a focus on applied studies and select study fields mostly linked to non-STEM occupations.
These are occupations in which frontier workers are underrepresented and complementarities
between native and foreign workers could arise. Our results are driven by individuals with a
vocational background at the upper-secondary level who have viable labor market options.
Downloaded from by guest on September 4, 2025. Copyright 2023

We show that their enrollment decisions are consistent with a rising wage premium for
tertiary non-STEM degrees.
The education system in Switzerland, similar to other European countries, grants access
to tertiary education to individuals with a general and a vocational background at the upper-
secondary level. At the tertiary level, they usually enroll at different institutions with a focus
on general or specific skills, respectively. This institutional feature contributes to a labor
force with a diverse skill set. As we have shown, access to education for people with different
backgrounds provides an important margin to respond to changes in labor market conditions.
The opportunity to upgrade skills can facilitate the transitions between work and education
as a response to labor market shocks.
Immigrants often bring distinct skills to the labor market that complement native work-
ers. A sudden outflow of foreigners due to a more restrictive migration policy or deterio-
rating relative economic conditions in the host country could result in a shortage of skills
that foreign workers were previously supplying. This problem is exacerbated by a further
specialization among the native population due to changes in education demand as shown
in this study. As skill acquisition is typically a long-term process, these findings should be
taken into account when considering changes to immigration policies.

30
 Tables

Table 1: Characteristics of Native versus Cross-Border Workers

Native workers Cross-border commuters

Mean Sd Mean Sd
Age 40.508 1.278 40.258 1.899
Share women 0.374 0.055 0.295 0.086
Share with lower-secondary education 0.111 0.045 0.263 0.196
Share with upper-secondary education 0.643 0.071 0.495 0.145
Share with academic tertiary education 0.129 0.066 0.148 0.081
Share with professional tertiary education 0.116 0.028 0.094 0.050
Share in no managerial position 0.614 0.053 0.714 0.074
Share in junior or executive managerial position 0.217 0.038 0.203 0.053
Share in medium or senior managerial position 0.169 0.029 0.083 0.037
Share in STEM occupation 0.428 0.109 0.614 0.128
Share in non-STEM occupation 0.572 0.109 0.386 0.128
Downloaded from by guest on September 4, 2025. Copyright 2023

Source: SESS.
Note: The observation period is 1996–2016. Data is at the commuting zone level. Lower-secondary level of
education is compulsory education as highest degree; upper-secondary is a degree from an upper-secondary
education with or without a matura exam; academic tertiary is a degree from a university or university of
applied sciences; and professional tertiary is a degree from other types of higher education institutions. Ob-
servations are weighed by the number of native workers or cross-border commuters, respectively.

31
 Table 2: Exposure to Cross-Border Commuters by Educational Level

Outcome: share of cross-border commuters

All Up to Upper-secondary Tertiary
lower-secondary
(1) (2) (3) (4)
30min * 2002-2006 0.013** -0.002 0.021** 0.005
(0.006) (0.007) (0.008) (0.007)
30min * 2008-2016 0.033*** 0.014 0.046*** 0.032***
(0.012) (0.009) (0.016) (0.011)

Mean outcome 0.072 0.070 0.069 0.069

Sd outcome 0.109 0.129 0.103 0.098
Commuting zones 106 106 106 106
within 30 min 35 35 35 35
N 1166 1166 1166 1160

Source: SESS.
Note: The table shows difference-in-differences estimates using biennial data at the commuting zone level for
Downloaded from by guest on September 4, 2025. Copyright 2023

the period 1996–2016. The dependent variable is the share of cross-border commuters in total employment
by educational level. Observations are weighed by the number of total employees in 1996. Standard errors
in parentheses are clustered at the commuting zone level. * p<0.1; ** p<0.05; *** p<0.01.

32
 Table 3: Cross-Border Commuters Relative to Resident Workers by Study Field

Field of study STEM field Skill supply of commuters

relative to residents
(1) (2)
Education 0 0.495
Languages 0 0.596
Law 0 0.653
Welfare 0 0.663
Journalism and information 0 0.670
Personal services 0 0.719
Humanities (except languages) 0 0.728
Social and behavioral sciences 0 0.764
Health 1 0.800
Veterinary 0 0.819
Business and administration 0 0.883
Arts 0 1.179
Mathematics and statistics 1 1.318
Biological and related sciences 1 1.384
Agriculture 1 1.547
Downloaded from by guest on September 4, 2025. Copyright 2023

Manufacturing and processing 1 1.549

Environment 1 1.613
Physical sciences 1 1.652
Engineering and engineering trades 1 1.948
Forestry 1 1.968
Information and communication technologies (ICT) 1 2.304
Architecture and construction 1 2.470

Sources: EHA, FSO.

Note: Column 1 distinguishes between STEM and non-STEM two-digit ISCED study fields. Column 2
shows the ratio of the share of commuters trained in a study field relative to the share of residents trained
in the same field according to Equation 4.

33
 Table 4: Exposure to Cross-Border Commuters by Education and Occupation

Outcome: share of cross-border commuters

Upper-secondary Tertiary
STEM Non-STEM STEM Non-STEM
(1) (2) (3) (4)
30min * 2002-2006 0.024*** 0.008 0.023 0.004
(0.009) (0.006) (0.014) (0.007)
30min * 2008-2010 0.033** 0.019* 0.044** 0.026**
(0.015) (0.011) (0.020) (0.010)

Mean outcome 0.081 0.039 0.096 0.050

Sd outcome 0.125 0.063 0.129 0.069
Commuting zones 106 106 106 106
within 30 min 35 35 35 35
N 848 848 824 837

Source: SESS.
Downloaded from by guest on September 4, 2025. Copyright 2023

Note: The table shows difference-in-differences estimates using biennial data at the commuting zone level for
the period 1996–2010. The dependent variable is the share of cross-border commuters in total employment
by educational level and occupation. Observations are weighed by the number of total employees in 1996.
Standard errors in parentheses are clustered at the commuting zone level. * p<0.1; ** p<0.05; *** p<0.01.

34
 Table 5: Summary Statistics

Affected areas Non-affected areas

N Mean Sd N Mean Sd
Enrollment rate 735 0.354 0.094 1491 0.311 0.083
... at universities 735 0.206 0.082 1491 0.170 0.060
... at universities of applied sciences 735 0.148 0.053 1491 0.141 0.049
... in agriculture 735 0.003 0.002 1491 0.004 0.003
... in arts and humanities 735 0.039 0.017 1491 0.031 0.014
... in business and law 735 0.092 0.028 1491 0.081 0.026
... in engineering 735 0.050 0.015 1491 0.048 0.014
... in health and welfare 735 0.046 0.028 1491 0.035 0.022
... in ICT 735 0.011 0.006 1491 0.011 0.006
... in math and sciences 735 0.033 0.012 1491 0.029 0.011
... in services 735 0.004 0.005 1491 0.003 0.004
... in social sciences 735 0.038 0.020 1491 0.031 0.015

Mean ln gross hourly wage 385 3.573 0.098 781 3.564 0.109
... of lower-secondary educated 385 3.298 0.082 781 3.297 0.083
... of upper-secondary educated 385 3.519 0.081 781 3.496 0.081
Downloaded from by guest on September 4, 2025. Copyright 2023

... of tertiary educated 385 3.934 0.088 774 3.937 0.085

Share employed in management 385 0.144 0.031 781 0.141 0.029

... with lower-secondary education 385 0.027 0.023 780 0.025 0.023
... with upper-secondary education 385 0.107 0.026 781 0.102 0.025
... with tertiary education 385 0.439 0.093 774 0.439 0.097

ln number employed 385 10.094 0.908 781 10.347 1.188

... with lower-secondary education 385 7.961 0.859 781 7.971 0.976
... with upper-secondary education 385 9.607 0.861 781 9.888 1.132
... with tertiary education 385 8.333 1.198 774 8.724 1.567

Sources: SESS, SHIS-studex.

Note: The table shows summary statistics for native outcome variables. The observation period for the en-
rollment outcomes is 1997–2017 and for the labor market outcomes 1996–2016. Data is at the commuting
zone level. Lower-secondary level of education is compulsory education as highest degree; upper-secondary
is a degree from an upper-secondary education with or without a matura exam; tertiary is a degree from a
university or university of applied sciences. Native enrollment rate is the share of native first-year students
in the birth cohort. One-digit ISCED fields of studies are considered. Observations are weighed by the na-
tive cohort size in 1997 and the number of native employees in 1996 by education group, respectively.

35
 Table 6: Native Enrollment Rate by Institutional Type

Outcome: share of native first-year students in birth cohort

All University University of applied
sciences
(1) (2) (3)
30min * 2002-2006 -0.000 -0.003 0.003
(0.007) (0.004) (0.004)
30min * 2007-2017 0.010 -0.001 0.011**
(0.007) (0.006) (0.004)

Mean outcome 0.326 0.183 0.143

Sd outcome 0.089 0.071 0.050
Commuting zones 106 106 106
within 30 min 35 35 35
N 2226 2226 2226

Source: SHIS-studex.
Note: The table shows difference-in-differences estimates using annual data at the commuting zone level for
Downloaded from by guest on September 4, 2025. Copyright 2023

the period 1997–2017. The dependent variable is the share of native first-year students in the birth cohort by
institutional type. Observations are weighed by the cohort size in 1997. Standard errors in parentheses are
clustered at the commuting zone level. * p<0.1; ** p<0.05; *** p<0.01.

36
 Table 7: Native Enrollment Rate by Type of Study Field

Outcome: share of native first-year students in birth cohort

STEM Non-STEM Affected Non-affected
(1) (2) (3) (4)
Panel A: All institutions
30min * 2002-2006 0.000 -0.001 -0.001 -0.000
(0.004) (0.005) (0.003) (0.005)
30min * 2007-2017 0.002 0.005 0.001 0.007
(0.003) (0.005) (0.003) (0.005)
Mean outcome 0.119 0.206 0.103 0.222
Sd outcome 0.035 0.061 0.025 0.071
Commuting zones 106 106 106 106
within 30 min 35 35 35 35
N 2226 2226 2226 2226
Panel B: Universities
Downloaded from by guest on September 4, 2025. Copyright 2023

30min * 2002-2006 -0.001 -0.003 -0.001 -0.002

(0.002) (0.003) (0.002) (0.003)
30min * 2007-2017 -0.000 -0.003 -0.002 -0.001
(0.002) (0.004) (0.002) (0.004)
Mean outcome 0.066 0.116 0.051 0.131
Sd outcome 0.026 0.049 0.019 0.057
Commuting zones 106 106 106 106
within 30 min 35 35 35 35
N 2226 2226 2226 2226
Panel C: Universities of applied sciences
30min * 2002-2006 0.001 0.001 0.001 0.002
(0.003) (0.003) (0.002) (0.004)
30min * 2007-2017 0.003 0.008*** 0.003 0.008**
(0.003) (0.003) (0.002) (0.003)
Mean outcome 0.053 0.091 0.052 0.091
Sd outcome 0.020 0.039 0.014 0.043
Commuting zones 106 106 106 106
within 30 min 35 35 35 35
N 2226 2226 2226 2226

Source: SHIS-studex.
Note: The table shows difference-in-differences estimates using annual data at the commuting zone level for
the period 1997–2017. Affected fields are those with a relative skill supply measure above one as shown in
Table 3. The dependent variable is the share of native first-year students in the birth cohort by study field
and institutional type. Observations are weighed by the cohort size in 1997. Standard errors in parentheses
are clustered at the commuting zone level. * p<0.1; ** p<0.05; *** p<0.01.

37
 Table 8: Native Labor Market Outcomes by Educational Level

All Up to Upper-secondary Tertiary

lower-secondary
(1) (2) (3) (4)

Panel A: ln gross hourly wage rate of natives

30min * 2002-2006 -0.003 -0.004 -0.005 0.019**

(0.005) (0.009) (0.006) (0.009)
30min * 2008-2010 -0.005 -0.003 -0.009 0.029***
(0.006) (0.013) (0.007) (0.010)
30min * 2012-2016 -0.014*** -0.012 -0.016*** 0.019*
(0.005) (0.017) (0.005) (0.011)
Mean outcome -0.000 -0.008 0.001 0.001
Sd outcome 0.060 0.074 0.061 0.074
Commuting zones 106 106 106 105
within 30 min 35 35 35 35
N 1166 1166 1166 1144
Downloaded from by guest on September 4, 2025. Copyright 2023

Panel B: Share of natives in a managerial position

30min * 2002-2006 0.005* -0.001 0.001 0.029

(0.003) (0.004) (0.004) (0.020)
30min * 2008-2010 0.003 0.007 -0.002 0.023
(0.004) (0.005) (0.005) (0.020)
30min * 2012-2016 0.002 -0.002 -0.001 0.022
(0.004) (0.004) (0.004) (0.026)
Mean outcome -0.000 -0.001 -0.000 -0.002
Sd outcome 0.023 0.024 0.024 0.093
Commuting zones 106 106 106 105
within 30 min 35 35 35 35
N 1166 1166 1166 1144
Panel C: ln number of natives employed

30min * 2002-2006 0.008 0.012 -0.018 0.088*

(0.033) (0.066) (0.040) (0.053)
30min * 2008-2010 0.013 0.041 -0.000 -0.015
(0.047) (0.088) (0.055) (0.078)
30min * 2012-2016 -0.051 0.033 -0.039 -0.107
(0.069) (0.061) (0.086) (0.111)
Mean outcome 9.553 7.011 9.112 7.952
Sd outcome 1.136 1.058 1.073 1.505
Commuting zones 106 106 106 105
within 30 min 35 35 35 35
N 1166 1166 1166 1144

Source: SESS.
Note: The table shows difference-in-differences estimates using biennial data at the commuting zone level for
the period 1996–2016. The sample consists of employees aged 18–40. The dependent variable in Panel A is
the mean natural log of gross hourly wage of natives (residualized) in an education category, in Panel B the
share of natives holding at least a middle management position (residualized) in an education category, and
in Panel C the natural log of number of natives employed in education category. Observations are weighed
by the number of native employees in a specific education category in 1996. Standard errors in parentheses
are clustered at the commuting zone level. * p<0.1; ** p<0.05; *** p<0.01.

38
 Table 9: Native Labor Market Outcomes by Education and Occupation

Upper-secondary Tertiary
STEM Non-STEM STEM Non-STEM
(1) (2) (3) (4)

Panel A: ln gross hourly wage rate of natives

30min * 2002-2006 0.003 -0.008 0.019 0.022

(0.005) (0.006) (0.012) (0.021)
30min * 2008-2010 0.004 -0.009 0.012 0.041
(0.006) (0.007) (0.016) (0.026)
Mean outcome 0.001 0.001 0.003 -0.001
Sd outcome 0.052 0.079 0.070 0.089
Commuting zones 106 106 98 100
within 30 min 35 35 33 35
N 848 848 766 791
Panel B: Share of natives in a managerial position
Downloaded from by guest on September 4, 2025. Copyright 2023

30min * 2002-2006 0.004 -0.006 -0.020 0.053***

(0.006) (0.005) (0.031) (0.016)
30min * 2008-2010 0.002 -0.010 -0.032 0.032
(0.007) (0.006) (0.034) (0.022)
Mean outcome -0.000 0.000 -0.007 -0.001
Sd outcome 0.028 0.032 0.130 0.113
Commuting zones 106 106 98 101
within 30 min 35 35 33 35
N 848 848 766 798
Panel C: ln number of natives employed

30min * 2002-2006 -0.049 0.012 0.047 0.097

(0.048) (0.048) (0.078) (0.083)
30min * 2008-2010 -0.085 0.057 -0.138 0.077
(0.055) (0.068) (0.090) (0.118)
Mean outcome 8.059 8.670 6.726 7.407
Sd outcome 0.908 1.224 1.393 1.598
Commuting zones 106 106 98 101
within 30 min 35 35 33 35
N 848 848 766 798

Source: SESS.
Note: The table shows difference-in-differences estimates using biennial data at the commuting zone level for
the period 1996–2010. The sample consists of employees aged 18–40. The dependent variable in Panel A is
the mean natural log of gross hourly wage of natives (residualized) in an education-occupation category, in
Panel B the share of natives holding at least a middle management position (residualized) in an education-
occupation category, and in Panel C the natural log of number of natives employed in education-occupation
category. Observations are weighed by the number of upper-secondary educated native employees in 1996 in
Columns 1–2 and tertiary educated native employees in 1996 in Columns 3–4. Standard errors in parentheses
are clustered at the commuting zone level. * p<0.1; ** p<0.05; *** p<0.01.

39
 Table 10: Native Enrollment Rate at Universities of Applied Sciences by Educational Back-
ground

Outcome: share of native first-year students in birth cohort

General Vocational Vocational Vocational Other
STEM non-STEM
(1) (2) (3) (4) (5)
30min * 2002-2006 -0.002 0.003 -0.001 0.004* 0.002
(0.002) (0.003) (0.002) (0.002) (0.001)
30min * 2007-2017 -0.003 0.013*** 0.003 0.009*** 0.001
(0.002) (0.004) (0.002) (0.003) (0.001)

Mean outcome 0.033 0.106 0.042 0.059 0.004

Sd outcome 0.018 0.040 0.017 0.031 0.005
Commuting zones 106 106 106 106 106
within 30 min 35 35 35 35 35
N 2226 2226 2226 2226 2226

Source: SHIS-studex.
Downloaded from by guest on September 4, 2025. Copyright 2023

Note: The table shows difference-in-differences estimates using annual data at the commuting zone level for the
period 1997–2017. The dependent variable is the share of native first-year students in the birth cohort. Col-
umn 1 shows first-year students with a general education (general matura), Column 2 first-year students with a
vocational background (vocational or specialized education). In Columns 3 and 4 we split the individuals with
a vocational background into STEM and non-STEM. Column 5 shows results for first-year students who can-
not be classified as generally or vocationally educated. Observations are weighed by the cohort size in 1997.
Standard errors in parentheses are clustered at the commuting zone level. * p<0.1; ** p<0.05; *** p<0.01.

40
 Figures
Downloaded from by guest on September 4, 2025. Copyright 2023

University
University of Applied Sciences

Figure 1: Tertiary Institutions in Affected and Non-Affected Areas in 2017

Source: Swisstopo.
Note: The map shows Switzerland’s 106 commuting zones split into affected (grey) and non-affected units
(white) and the location of tertiary institutions by institutional type.

41
 (a) 1997 (b) 2017
Downloaded from by guest on September 4, 2025. Copyright 2023

Figure 2: Share of Cross-Border Commuters and Travel Time

Source: FSO.
Note: The figure shows estimates from a locally weighted regression of the share of cross-border commuters
in 1997 in Subfigure (a) and 2017 in Subfigure (b) relative to 1995 employment on travel time to the closest
Swiss border crossing. The unit of observation is the commuting zone. The dashed line plots the function
exp(−0.05 × travel time) rescaled by ten in Subfigure (a) and five in Subfigure (b). The vertical line is
drawn at 30 minutes travel time.

42
 .1
(a) All cross-border commuters (b) Up to lower-secondary educated

.1
.08

.08
.06

.06
.04

.04
.02

.02
0

0
-.02

-.02
-.04

-.04
1996 2000 2004 2008 2012 2016 1996 2000 2004 2008 2012 2016

(c) Upper-secondary educated (d) Tertiary educated

Downloaded from by guest on September 4, 2025. Copyright 2023

.1

.1
.08

.08
.06

.06
.04

.04
.02

.02
0

0
-.02

-.02
-.04

-.04

1996 2000 2004 2008 2012 2016 1996 2000 2004 2008 2012 2016

Figure 3: Exposure to Cross-Border Commuters by Educational Level

Source: SESS.
Note: The figure shows difference-in-differences estimates using biennial data at the commuting zone level for
the period 1996–2016. The reference year is 2000. The vertical lines indicate the beginning of the transition
period (2002) and of the post-reform period (2007). The dependent variable is the share of cross-border
commuters in total employment by educational level. Observations are weighed by the number of total
employees in 1996. Standard errors are clustered at the commuting zone level; 95% confidence intervals are
shown.

43
 (a) All

.06
.04
.02
0
-.02
-.04

1990 1994 1998 2002 2006 2010 2014 2018

(b) University (c) University of applied sciences

Downloaded from by guest on September 4, 2025. Copyright 2023

.06

.06
.04

.04
.02

.02
0

0
-.02

-.02
-.04

-.04

1990 1994 1998 2002 2006 2010 2014 2018 1990 1994 1998 2002 2006 2010 2014 2018

Figure 4: Native Enrollment Rate by Institutional Type

Source: SHIS-studex.
Note: The figure shows difference-in-differences estimates using annual data at the commuting zone level for
the period 1991–2017. The reference year is 2001. The vertical lines indicate the beginning of the transition
period (2002) and of the post-reform period (2007). The dependent variable is the share of native first-year
students in the birth cohort by institutional type. Observations are weighed by the cohort size in 1997.
Standard errors are clustered at the commuting zone level; 95% confidence intervals are shown.

44
 .04
(a) STEM (b) Non-STEM

.04
.02

.02
0

0
-.02

-.02
1996 2000 2004 2008 2012 2016 1996 2000 2004 2008 2012 2016
Downloaded from by guest on September 4, 2025. Copyright 2023

Figure 5: Native Enrollment Rate by Type of Study Field at Universities of Applied Sciences
Source: SHIS-studex.
Note: The figure shows difference-in-differences estimates using annual data at the commuting zone level for
the period 1997–2017. The reference year is 2001. The vertical lines indicate the beginning of the transition
period (2002) and of the post-reform period (2007). The dependent variable is the share of native first-year
students in the birth cohort by study field. Observations are weighed by the cohort size in 1997. Standard
errors are clustered at the commuting zone level; 95% confidence intervals are shown.

45
 References

Abramitzky, Ran, Victor Lavy, and Maayan Segev. forthcoming. “The Effect of
Changes in the Skill Premium on College Degree Attainment and the Choice of Major.”
Journal of Labor Economics.

Acton, Riley K. 2021. “Community College Program Choices in the Wake of Local Job
Losses.” Journal of Labor Economics, 39(4): 1129–1154.

Anelli, Massimo, Kevin Shih, and Kevin Williams. forthcoming. “Foreign Students
Downloaded from by guest on September 4, 2025. Copyright 2023

in College and the Supply of STEM Graduates.” Journal of Labor Economics.

Atkin, David. 2016. “Endogenous Skill Acquisition and Export Manufacturing in Mexico.”
American Economic Review, 106(8): 2046–85.

Ayllón, Sara, and Natalia Nollenberger. 2021. “The Unequal Opportunity For Skills
Acquisition During the Great Recession In Europe.” Review of Income and Wealth, 67(2):
289–316.

Bartik, Timothy. 1991. Who Benefits from State and Local Economic Development Poli-
cies?.: W.E. Upjohn Institute for Employment Research.

Beerli, Andreas, Jan Ruffner, Michael Siegenthaler, and Giovanni Peri. 2021.
“The Abolition of Immigration Restrictions and the Performance of Firms and Workers:
Evidence from Switzerland.” American Economic Review, 111(3): 976–1012.

Beffy, Magali, Denis Fougre, and Arnaud Maurel. 2012. “Choosing the Field of Study
in Postsecondary Education: Do Expected Earnings Matter?.” The Review of Economics
and Statistics, 94(1): 334–347.

Betts, Julian. 1998. “Educational Crowding Out: Do Immigrants Affect the Educational
Attainment of American Minorities?.” In Help or Hindrance? The Economic Implications

46
 of Immigration for African Americans. eds. by Daniel S. Hamermesh, and Frank D. Bean,
New York: Russell Sage Foundation, 253–281.

Borjas, George J. 1995. “The Economic Benefits from Immigration.” Journal of Economic
Perspectives, 9(2): 3–22.

Borjas, George J. 2003. “The Labor Demand Curve is Downward Sloping: Reexamining
the Impact of Immigration on the Labor Market.” The Quarterly Journal of Economics,
118(4): 1335–1374.
Downloaded from by guest on September 4, 2025. Copyright 2023

Borjas, George J. 2004. “Do Foreign Students Crowd Out Native Students from Graduate
Programs?.” Working Paper 10349, National Bureau of Economic Research.

Borjas, George J. 2006. “Native Internal Migration and the Labor Market Impact of
Immigration.” Journal of Human Resources, 41(2): 221–258.

Borjas, George J., and Kirk B. Doran. 2012. “The Collapse of the Soviet Union and the
Productivity of American Mathematicians.” The Quarterly Journal of Economics, 127(3):
1143–1203.

Bundesamt für Statistik. 2018. 15 Bildung und Wissenschaft. Übergänge nach Abschluss
der Sekundarstufe II und Integration in den Arbeitsmarkt. Neuchâtel: Bundesamt für
Statistik.

Bundesamt für Statistik. 2019a. “Ausbildungswahl auf der Sekundarstufe II – Daten des
Indikators.”

Bundesamt für Statistik. 2019b. “Quote der Erstabschlüsse auf der Sekundarstufe II und
Maturitätsquote: Tabellen für das Jahr 2019.”

Bundesamt für Statistik. 2020. Soziale und wirtschaftliche Lage der Studierenden.
Neuchâtel: Bundesamt für Statistik.

47
 Callaway, Brantly, Andrew Goodman-Bacon, and Pedro H. C. Sant’Anna. 2021.
“Difference-in-Differences with a Continuous Treatment.”Technical Report 2107.02637,
arXiv.

Cortés, Patricia, and Jessica Pan. 2015. “Immigration and Occupational Choice of
Natives: the Case of Nurses in the United States.” CESifo Economic Studies, 61(3-4):
797–823.

Cristelli, Gabriele, and Francesco Lissoni. 2020. “Free Movement of Inventors: Open-
Border Policy and Innovation in Switzerland.” Munich Personal RePEc Archive 104120,
Downloaded from by guest on September 4, 2025. Copyright 2023

MPRA.

Dellas, Harris, and Plutarchos Sakellaris. 2003. “On the Cyclicality of Schooling: The-
ory and Evidence.” Oxford Economic Papers, 55(1): 148–172.

Denzler, Stefan, and Stefan C. Wolter. 2010. “Der Einfluss des lokalen Hochschulange-
bots auf die Studienwahl.” Zeitschrift für Erziehungswissenschaft, 13(4): 683–706.

Dustmann, Christian, Tommaso Frattini, and Ian P. Preston. 2012. “The Effect of
Immigration along the Distribution of Wages.” The Review of Economic Studies, 80(1):
145–173.

Dustmann, Christian, Uta Schönberg, and Jan Stuhler. 2016. “The Impact of Immi-
gration: Why Do Studies Reach Such Different Results?.” Journal of Economic Perspec-
tives, 30(4): 31–56.

Dustmann, Christian, Uta Schönberg, and Jan Stuhler. 2017. “Labor Supply Shocks,
Native Wages, and the Adjustment of Local Employment.” The Quarterly Journal of
Economics, 132(1): 435–483.

Edo, Anthony. 2019. “The Impact of Immigration on Wage Dynamics: Evidence from

48
 the Algerian Independence War.” Journal of the European Economic Association, 18(6):
3210–3260.

Goldsmith-Pinkham, Paul, Isaac Sorkin, and Henry Swift. 2020. “Bartik Instru-
ments: What, When, Why, and How.” American Economic Review, 110(8): 2586–2624.

Hanson, Gordon H., and Matthew J. Slaughter. 2018. “12. High-Skilled Immigration
and the Rise of STEM Occupations in US Employment.” In Education, Skills, and Tech-
nical Change. Implications for Future US GDP Growth. eds. by Charles R. Hulten, and
Valerie A. Ramey, Chicago: University of Chicago Press, 465–500.
Downloaded from by guest on September 4, 2025. Copyright 2023

Hoxby, Caroline M. 2009. “The Changing Selectivity of American Colleges.” Journal of

Economic Perspectives, 23(4): 95–118.

Hunt, Jennifer. 2017. “The Impact of Immigration on the Educational Attainment of

Natives.” Journal of Human Resources, 52(4): 1060–1118.

Hunt, Jennifer, and Marjolaine Gauthier-Loiselle. 2010. “How Much Does Immigra-
tion Boost Innovation?.” American Economic Journal: Macroeconomics, 2(2): 31–56.

Johnson, Matthew T. 2013. “The Impact of Business Cycle Fluctuations on Graduate

School Enrollment.” Economics of Education Review, 34(C): 122–134.

Llull, Joan. 2018. “Immigration, Wages, and Education: A Labour Market Equilibrium
Structural Model.” The Review of Economic Studies, 85(3): 1852–1896.

Long, Bridget Terry. 2014. The Financial Crisis and College Enrollment: How Have
Students and Their Families Responded?. 209–233, Chicago: University of Chicago Press.

Long, Mark C., Dan Goldhaber, and Nick Huntington-Klein. 2015. “Do completed
college majors respond to changes in wages?” Economics of Education Review, 49 1–14.

Machin, Stephen, and Richard Murphy. 2017. “Paying Out and Crowding Out? The
Globalization of Higher Education.” Journal of Economic Geography, 17(5): 1075–1110.

49
 Malamud, Ofer, and Abigail Wozniak. 2012. “The Impact of College on Migration:
Evidence from the Vietnam Generation.” Journal of Human Resources, 47(4): 913–950.

Manning, Alan, and Barbara Petrongolo. 2017. “How Local Are Labor Markets? Evi-
dence from a Spatial Job Search Model.” American Economic Review, 107(10): 2877–2907.

McHenry, Peter. 2015. “Immigration and the Human Capital of Natives.” Journal of
Human Resources, 50(1): 34–71.

Moser, Petra, Alessandra Voena, and Fabian Waldinger. 2014. “German Jewish
Émigrés and US Invention.” American Economic Review, 104(10): 3222–55.
Downloaded from by guest on September 4, 2025. Copyright 2023

OECD. 2022. “Education at a Glance.” Accessed 10.10.2022.

Orrenius, Pia M, and Madeline Zavodny. 2015. “Does Immigration Affect Whether
US Natives Major in Science and Engineering?.” Journal of Labor Economics, 33(S1):
S79–S108.

Ottaviano, Gianmarco, and Giovanni Peri. 2012. “Rethinking the Effect of Immigration
on Wages.” Journal of the European Economic Association, 10(1): 152–197.

Patnaik, Arpita, Matthew J Wiswall, and Basit Zafar. 2021. “College Majors.” In
The Routledge Handbook of the Economics of Education. ed. by Brian P. McCall, New
York: Routledge, 415–457.

Peri, Giovanni, Kevin Shih, and Chad Sparber. 2015. “STEM Workers, H-1B Visas,
and Productivity in US Cities.” Journal of Labor Economics, 33(S1): S225 – S255.

Peri, Giovanni, and Chad Sparber. 2009. “Task Specialization, Immigration, and
Wages.” American Economic Journal: Applied Economics, 1(3): 135–69.

Peri, Giovanni, and Chad Sparber. 2011. “Highly Educated Immigrants and Native
Occupational Choice.” Industrial Relations: A Journal of Economy and Society, 50(3):
385–411.

50
 Peri, Giovanni, and Vasil Yasenov. 2019. “The Labor Market Effects of a Refugee Wave.
Synthetic Control Method Meets the Mariel Boatlift.” Journal of Human Resources, 54(2):
267–309.

Ransom, Tyler, and John V Winters. 2021. “Do foreigners crowd natives out of STEM
Degrees and occupations? Evidence from the US Immigration Act of 1990.” ILR Review,
74(2): 321–351.

Schündeln, Matthias. 2014. “Are Immigrants More Mobile than Natives? Evidence from
Germany.” Journal of Regional Science, 54(1): 70–95.
Downloaded from by guest on September 4, 2025. Copyright 2023

Schweri, Jürg, and Joop Hartog. 2017. “Do Wage Expectations Predict College En-
rollment? Evidence from Healthcare.” Journal of Economic Behavior & Organization,
141(C): 135–150.

Shih, Kevin. 2017. “Do International Students Crowd-out or Cross-subsidize Americans in

Higher Education?.” Journal of Public Economics, 156(C): 170–184.

Wiswall, Matthew, and Basit Zafar. 2015. “Determinants of College Major Choice:
Identification using an Information Experiment.” The Review of Economic Studies, 82(2):
791–824.

Xia, Xiaoyu. 2016. “Forming Wage Expectations through Learning: Evidence from College
Major Choices.” Journal of Economic Behavior & Organization, 132(PA): 176–196.

51