“Too Shocked to Search”: COVID-19 and Apprenticeships
Summary
"Too shocked to search": The COVID-19 shutdowns' impact on the search for apprenticeships is an open-access research article by Daniel Goller and Stefan C. Wolter in the Swiss Journal of Economics and Statistics (2021) 157:6. It uses daily search queries on the national Swiss platform for apprenticeship vacancies from February 2020 until April 2021 as a proxy for the supply of potential apprentices. Analysing over 10 million search queries, the authors report a decrease of up to 40% during the first shutdown followed by a catch-up effect, and find that the link between the stringency of restrictions and search intensity diminished over time. The article reviews related literature, describes the Swiss apprenticeship market in 2020 with a table of open positions and signed contracts, and sets out its methodology and results.
Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used
Full text
Goller and Wolter Swiss J Economics Statistics (2021) 157:6 Swiss Journal of
https://doi.org/10.1186/s41937-021-00075-z
Economics and Statistics
ORIGINAL ARTICLE Open Access
“Too shocked to search” The COVID‑19
shutdowns’ impact on the search
for apprenticeships
Daniel Goller1,5* and Stefan C. Wolter1,2,3,4
Abstract
Even though the recession in Switzerland triggered by COVID-19 ultimately remained without consequences for
the apprenticeship market, significantly fewer apprenticeship contracts had been signed in the months of the first
shutdown in 2020 than in the same months of the previous year. Using daily search queries on the national adminis-
trative platform for apprenticeship vacancies from February 2020 until April 2021 as a proxy for the supply of potential
apprentices, we find a temporal pattern that coincides perfectly with the development of signed apprenticeship con-
tracts. Furthermore, the analyses show that the initially very strong relationship between the intensity of the politically
imposed restrictions to fight the COVID-19 pandemic and the daily search queries diminished over time, leading to a
search intensity in March 2021 that was back at pre-pandemic level.
Keywords: COVID-19, Switzerland, Stringency Index, Apprenticeship, Vacancies
JEL Classification: I20, J22
1 Motivation with school-based types of education (for an overview on
Switzerland is proud of its well-functioning vocational the effects of business cycles on apprenticeships see e.g.,
education and training system (VET). Training young Brunello, 2009 or Muehlemann & Wolter, 2021 and for
people specifically for the demands of a profession, as a general overview on the economics of apprenticeship
well as the needs of a company, has proven to be a good see e.g., Wolter & Ryan, 2011 or Muehlemann & Wolter,
remedy against youth unemployment and counteracts 2020). Basically, the apprenticeship market consists of
the shortage of skilled workers (e.g., Bolli et al., 2021). two sides, firms on the one side, demanding apprentices
Furthermore, apprentices benefit from the system as well, and on the other side the supply of young people willing
as they earn theoretical knowledge and working experi- to learn an occupation. The total number of training con-
ence while they are trained to become professionals in tracts therefore reflects the equilibrium of the number of
the respective job. training positions offered by firms willing to train, as well
However, all these advantages are also counterbal- as the number of individuals willing to learn a job.
anced by the potential disadvantage that the demand for This work investigates the supply of apprentices dur-
apprentices depends on the economic situation and does ing the COVID-19 pandemic as the potential reason for
not simply follow the demographic trend of young peo- a temporal decrease in signed training contracts in the
ple looking for post-compulsory education, as is the case first months of the pandemic. Using an innovative data
source, we are able to track the supply side in form of
revealed preferences on a daily basis by the number of
*Correspondence: Daniel.Goller@vwi.unibe.ch
1
search queries on the national Swiss platform for appren-
Centre for Research in Economics of Education, University of Bern, Bern,
Switzerland
ticeship vacancies. Although this does not show the
Full list of author information is available at the end of the article supply of apprenticeship seekers as such, it does show
© The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which
permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the
original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or
other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line
to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this
licence, visit http://creativecommons.org/licenses/by/4.0/.
Goller and Wolter Swiss J Economics Statistics (2021) 157:6 Page 2 of 15
the search intensity, which could not automatically be 2020 until February 2021 on the intensity of searching
derived from the number of registered apprenticeship for an apprenticeship. The empirical results show that
seekers. the temporal pattern of the progression of the cumulative
In the past it was only possible to make assumptions number of signed apprenticeship contracts is very simi-
as to whether it was supply or demand that primar- lar to the pattern of the search intensity for apprentice-
ily determined the number of apprenticeship contracts ship positions. Even though it is not possible to postulate
concluded, due to the very limited availability of data causal relationships based on the available data, the anal-
on supply and demand or even the complete absence of yses show that the development of the apprenticeship
such data, as is the situation in Switzerland. In the case market in Switzerland in the short run and in the wake of
of large demographic fluctuations, it was assumed, ceteris the COVID-19 pandemic was probably determined by a
paribus, that changes in contracts were due to the influ- reaction of supply.
ence of supply and, conversely, in the case of economic The rest of this work is structured as follows. We pro-
fluctuations, that the influence was due to changes in the vide a short review of emerging literature on COVID-19
demand for apprentices. Thus, this study contributes to effects in general and the effects on the labour market
a better understanding of the functioning of the appren- and job search behaviour, in Sect. 2. Section 3 provides
ticeship market in Switzerland, as for the first time the a short overview of the apprenticeship market in Swit-
supply of apprenticeship seekers can be tracked over zerland in 2020. Section 4 introduces the methodology,
time. data base and some descriptive statistics. Section 5 pre-
After the outbreak of the COVID-19 pandemic, there sents the results of the analysis, followed by concluding
were fears, based on experiences in previous crises remarks in Sect. 6.
(Luethi & Wolter, 2020), that the apprenticeship mar-
ket in Switzerland would primarily suffer as a result of 2 Literature
the expected economic recession and its impact on the There is a fast-growing COVID-19-related literature
demand for apprentices. In April and May, the number investigating economic consequences of the pandemic
of signed apprenticeship contracts fell significantly com- (compare e.g., Baker et al., 2020; Chetty et al., 2020;
pared to the previous year’s values. However, this num- Goolsbee & Syverson, 2021; among many others). For
ber returned to the previous year’s values later in the Switzerland, an array of papers was published document-
summer, so that when the apprenticeships of the 2020 ing the COVID-19 impact,2 often using very novel data
cohort began, the same number of apprentices started an available on an hourly, daily, weekly, or monthly base. The
apprenticeship as before the pandemic. As there was no outcomes included new measures of GDP (Burri & Kauf-
indication that the demand for apprentices changed sig- mann, 2020), mobility and sales (Eckert & Mikosch, 2020;
nificantly in the aftermath of the March 2020 shutdown Persson et al., 2021), trade (Buechel et al., 2020), or shifts
and it is unlikely that apprenticeship vacancies fluctu- in retail payments (Kraenzlin et al., 2020).
ate largely from month to month,1 we direct the focus There is also a growing literature on the effects of the
in this paper to the supply of potential apprentices as an pandemic on the labour market (e.g., Gupta et al., 2020;
explanation for the observed temporal pattern of signed Forsythe et al., 2020; Coibion et al., 2020; among others)
contracts. and more specifically on job search, which is a topic that
Analysing over 10 million search queries for appren- comes closest to this paper. For the Swiss labour market
ticeship positions from end of February 2020 until April Lalive et al. (2020) found that jobseekers during the cri-
2021, we find a sharp decrease of up to 40% in search sis invested less time in their job search than before the
queries during the first shutdown followed by a catch- Corona crisis. This was driven by a decline in the number
up effect afterwards. Further, we find that the effect of of vacancies, as well as due to being afraid of becoming
politically imposed restrictions on the search inten- infected with the virus during the recruitment process.
sity decreases over time. Consequently, we find a lower In another study using Dutch survey data, Balgova et al.
impact of the second shutdown starting in December (2021) found that unemployed individuals reduced their
effort in searching for a job, while employed individu-
als searched rather more intensively. Finally, Marinescu
1
Contrary to hiring regular workers, apprenticeships are long-term invest- et al. (2021) found job applications during the COVID-19
ments of (training) companies as they come with contracts that last either
three or four years. Therefore, for most companies the decision to offer train-
ing positions is a strategic one, followed by a time-consuming process of
advertising, interviewing, and selecting the right candidates. Bearing this in
2
mind, it is rather unlikely that interested companies would have withdrawn The Swiss Journal of Economics and Statistics, among other journals, started
their advertised vacancies only to put them again on the market some weeks a special focus with many more articles on the economic outcomes of the
later or replaced them by new ones in such a short time. COVID-19 pandemic in 2020.
Goller and Wolter Swiss J Economics Statistics (2021) 157:6 Page 3 of 15
Fig. 1 Difference in signed apprenticeship contracts compared to previous year. Differences in signed contracts compared to previous year by
NUTS-2 regions. The red solid line represents the Swiss average. Data source: “Taskforce Perspektive Berufslehre 2020”
pandemic to decrease. They further found this to be German statistics. The German studies have the possi-
related to increased unemployment benefits. This rela- bility to rely on proxies for the potential supply of new
tionship weakened as the number of job applications apprentices by using data on the registered applicants in
increased again starting in May 2020. the data base of the Federal Employment Agency. A data
The above-described implications and observations set that does not exist in Switzerland as apprenticeship
for the labour market in general can also be found in the seekers cannot register in an official register.
apprenticeship market, even though the conditions for
potential apprentices and the objectives of companies are 3 The apprenticeship market in Switzerland
different than in the case of regular workers. In an early in 2020
approach to predict the supply and demand side of the There are no official data sources in Switzerland that
German apprenticeship market in the first year of the would have allowed the supply of and demand for
COVID-19 crisis, Muehlemann et al. (2020) estimated, apprenticeships to be monitored periodically and at
in contrast to the Swiss market, a COVID-19 induced high frequency. However, because of the emerging eco-
6% reduction in signed apprenticeship positions in 2020 nomic consequences of the pandemic, the federal gov-
in Germany. Similar predictions with different data sets ernment set up a task force which, among other things,
and methods were made by Maier (2020) predicting a took on the regular monitoring of the situation on the
reduction in signed apprenticeship contracts in Ger- apprenticeship market. As one of these measures, the
many induced by a simultaneous decrease in supply and cantons reported monthly on the apprenticeship con-
demand and Oeynhausen et al. (2020) confirming, that tracts signed and compared these with the number of
indeed during the COVID-19 crisis there was a substan- contracts signed in the same month of the previous year.
tial decrease in firms’ demand for apprentices but also In this way, the situation on the apprenticeship market
applications sent to companies and consequently a reduc- could be monitored at least monthly, starting from April
tion also in the number of signed contracts in the official 2020 onwards.
Goller and Wolter Swiss J Economics Statistics (2021) 157:6 Page 4 of 15
Table 1 Open apprenticeships (total) and signed contracts (cumulative), 2020 and 2021. Source: “Taskforce Perspektive Berufslehre
2020”. Numbers are recorded at the end of the respective month. The monthly changes in the number of vacancies and signed
contracts are not equal, because the number of vacancies does not cover all vacancies, but only those posted on the national platform
2020 2021
Open positions Signed contracts Open positions Signed contracts
April 22,195 42,212 21,083 43,106
May 19,271 47,811 17,889 49,642
June 16,221 55,454 14,502 57,946
Looking at these figures (Fig. 1), one can see that the positions in 2020, and the share of companies that have
number of signed apprenticeship contracts declined in offered more or fewer had hardly changed. More spe-
the first half of the year 2020 relative to 2019, but also a cifically, the number of firms that had reduced their
catch-up effect reaching the same number of apprentice- numbers of positions was unchanged compared to the
ship contracts by August 2020 compared to the previ- previous years (2020: 8%; 2019: 8%; 2018: 9%).
ous year. The substantial differences in signed contracts Third, the SERIs “Taskforce Perspektive Berufslehre”
between “Nomenclature of Territorial Units for Statis- reported monthly from April 2020 onwards the signed
tics-2” (NUTS-2) regions can mainly be related to dif- contracts and open apprenticeship positions registered.
ferences in the timing of the conclusion of the contracts. In Table 1 we contrast the number of open positions and
Traditionally, contracts are already signed as early as signed contracts in the years 2020 to 2021. It is notewor-
October in the year prior to the start of the apprentice- thy that the number of vacancies in 2020 is very similar
ship in the German speaking cantons, whereas in the to the number in 2021 and the monthly changes in the
French and Italian speaking part of the country firms and numbers of vacancies and the number of signed con-
students wait till just a few months before the start of tracts show the same ratios in both years. While there is
the apprenticeship.3 Thus, while for the German-speak- not every apprenticeship offer registered we can be sure
ing cantons a major part of contracts was already signed that there was no lack of open positions, which could jus-
before the COVID-19 outbreak, this process just started tify the difference in signed contracts in Fig. 1. Fourth,
in Région lémanique and Ticino, which Fig. 5 in “Appen- the private provider Yousty issued a press release in April
dix” documents for April 2020. 2020 stating that the number of open positions barely
changed since the beginning of the COVID-19 outbreak.4
3.1 Demand for apprentices Therefore, as there are no signs that the demand for
Regarding the demand for apprentices alone, there are apprentices changed much in the aftermath of the shut-
concurrent indications that the number of offered posi- down in March 2020, the focus of this work is to investi-
tions to start the apprenticeship in 2020 did not change gate whether the supply side in the Swiss apprenticeship
substantially compared to previous years. First, as Fig. 1 market could explain this striking pattern of an initial
documents, the number of signed apprenticeship con- decline in signed apprenticeship contracts and subse-
tracts reached the previous years’ numbers by the end quent catch-up.
of August 2020. Thus, it is fair to assume that the total
demand for apprentices did not change. Second, the 3.2 Supply of apprentices
“Nahtstellenbarometer” of the SERI monitors the train- The supply side of prospective apprentices consists
ing companies twice a year in April and in August. These mainly of graduates of compulsory school, but also those
representative surveys showed that the firms’ demand entering the apprenticeship market after an interim solu-
for apprentices had not changed in 2020 relative to 2019. tion, youth re-orienting after having dropped out of
In August 2020, most of the companies reported, that general education or even those intending to train for
they had not changed their number of apprenticeship a second occupation after having already successfully
3 4
The latter is a consequence of the fact that most students prefer a general Sources are «Nahtstellenbarometer»: https://cockpit.gfsbern.ch/de/cockpit/
education and therefore wait to see whether they will gain access to this edu- nahtstellenbarometer-2020-3/; «Perspektive Berufslehre»: https://taskforce2
cation or not. Only if they do not make it do they look for an apprenticeship. 020.ch/de/monitoring; Yousty April 2020 press release: https://www.yousty.
ch/de-CH/Pressemitteilungen. (In German; accessed 23.07.2021).
Goller and Wolter Swiss J Economics Statistics (2021) 157:6 Page 5 of 15
completed an apprenticeship or a general education. For million search queries.6 Figure 2 shows how the number
them, applying to occupations became more difficult of daily search queries evolved over this time. The major
during the COVID-19 pandemic. advantage of our measure of apprentices supply, even
From the perspective of these individuals, the situation though we do not know the number of potential appren-
they were confronted with was unique in several aspects. tices, is that we observe revealed preferences on a daily
First, with the classification of the evolving COVID- basis, with those interested in apprenticeships sending
19 pandemic as “exceptional situation”, the daily life of search queries. Further, they stop searching as soon as
youth changed radically. It required significant sacrifice they found an apprenticeship or lost interest (also tem-
and acclimatisation, e.g., to a life of closed schools and porarily) for other reasons, which gives us a good proxy
distance learning. Second, there was great uncertainty for the daily interest in apprenticeship training posi-
about the length and intensity of this crisis. Applying for tions, by field of occupation and region. Other measures
vacant apprenticeship posts may not be advisable, as the are mainly only available on a yearly basis or as rough
firms might run into financial problems or might not be proxies.
able to provide a full vocational education and training The (weekly) average of daily queries one and two weeks
for other reasons. Third, as many businesses were forced before and after the first shutdown on March 16th, 2020,
to close their operations or work from home during the are shown in Table 2. We observe decreasing numbers of
shutdown, the application process became difficult and daily search queries in 2020 accompanied by an increasing
complicated, so much so that many potential applicants Stringency Index, a measure of toughness of the politically
tended not to apply at all. Fourth, although this was more imposed restrictions in response to the pandemic—which
of concern for the cohort looking for an apprenticeship is described in more detail in the following. For 2021 both,
in 2021, the usual training fairs and information days in daily queries and the Stringency Index are rather on a con-
the companies, as well as trial (apprenticeships) working stant level comparable to the “pre-crisis” level.
days could not take place. Young people were therefore For the further analysis, we aggregate the individual
not able to inform themselves about potential employ- queries by day, canton, and occupation category, lead-
ers and job profiles as was the practice in previous years. ing to a total of 512,148 observations.7 To measure the
Putting all this together, it is quite realistic to assume that effect of the shutdown policy on the number of search
the supply of learners has been affected by the COVID- queries we focus on two different types of variables of
19 pandemic. interest. First, a binary variable is constructed which
is equal to 1 starting with the national shutdown from
4 Data and methodology the 16th March, 2020 onwards until the (second) eas-
Our data base consists of every user query searching for ing of the national restrictions on the 11th May, 2020.8
an apprenticeship on the public national online appren- To investigate the second (national) shutdown a binary
ticeship platform from the 28th February 2020 until the variable is created which equals 1 for the days from the
04th April 2021.5 This official platform operates on behalf 22nd December 2020 until 28th February 2021. Second,
of the Swiss cantons and is not only the main source for
apprenticeship vacancies in Switzerland, but also the offi-
cial Swiss information portal for study and career guid- 6
From the initial 12 million search queries, we removed all search queries in
ance. People interested in starting an apprenticeship Liechtenstein. We removed those not specifying an occupation or location,
which is needed to investigate containment policies on a cantonal level. Fur-
can search in all four national languages by occupation ther, we removed all automated (crawler/bot) queries, leading to a final set of
and location for open apprenticeships. While not every about 10 million search queries. To set this number into perspective: the 10
apprenticeship position is advertised online and there are million queries result not only from individuals interested in an apprentice-
ship position. Parents, teachers, firms, administrations, and career guidance
other (private) online platforms, this is the most com- centres gather information there for advising the potential apprentices. Still,
prehensive and a well-known outlet for apprenticeship the search for a suitable apprenticeship is complex and requires searching
advertisements, guaranteeing representativeness of the several times on several days until a contract is signed or interest in training
is lost. During the COVID-19 crisis (no trial apprenticeships, no information
data. In the investigated 13 months we observe over 10 days, etc.) individuals searching for an apprenticeship were more dependent
than usual on the help of confidants and institutions. Thus, even though we
cannot distinguish the senders of search queries, if having a bias, we expect
5
berufsberatung.ch (de; fr: orientation.ch; it: orientamento.ch) is the official the estimated effect to be a conservative, lower bound effect.
platform for information about and searching for apprenticeships offered by 7
We chose to use the subgroups of the SwissDoc-number classification,
the SDBB | CFSO, which is an agency of the Swiss Conference of Cantonal resulting in 49 occupational groups. 402 (days) × 26 (cantons) × 49 (occupa-
Ministers of Education. Although there are also private platforms and there tions) = 512,148 observations.
are other ways to search for apprenticeships, the national platform that col- 8
To investigate if the effect is driven by the definition of this variable, as
lects data on vacancies in all cantons is probably the platform that most well as to see how the effect evolved over time, we chose to add various
apprenticeship seekers will consult during their search and application pro- other definitions, i.e., for 7, 14, 30, 80 and 100 days, as well as until the first
cess at least once. easing of the restriction (25th April, 2020).
Goller and Wolter Swiss J Economics Statistics (2021) 157:6 Page 6 of 15
Fig. 2 Total daily queries and Stringency Index. Green solid line represents the daily queries, Nadaraya-Watson Kernel-smoothed (left axis), the black
broken line is the Stringency Index (right axis), exemplarily for the canton Zurich
Table 2 Daily queries before and after the first shutdown, 2020 vs. 2021
2020 2021
Daily queries (7 days Stringency Index Daily queries (7 days Stringency Index
average) average)
March, 02.–08. 30,801 0.250 31,005 0.620
March, 09.–15. 26,971 0.349 30,936 0.620
March, 17.–23. 13,261 0.771 30,771 0.620
March, 24.–30. 15,164 0.787 29,035 0.620
Averages of daily queries in week 2 and week 1 before and after the date of the first shutdown, March 16th, 2020. The Stringency Index is a measure for restrictiveness
of politically imposed countermeasures, which is described later in this section
we use the “KOF Stringency Index/KOF Stringency Index The availability of the KOF Stringency Index on a can-
Plus” (Pleninger et al., 2021), which are constructed as tonal base gives us the advantage that we can capture not
composite measures of the different restrictions and are only the impact of variations of this index over time but
ranging from 0 (no restrictions) to 1 (full shutdown).9 also between geographical regions, as in Switzerland,
like in other countries, different levels of restrictions
9
were in function for different regions at the same time.
It combines 9 indicators to an equally weighted one-dimensional index:
school closures, workspace closures, cancellation of public events, closure of The construction of the daily KOF Stringency Indices fol-
public transport, restrictions on gatherings, requirements to stay at home, lows closely the Oxford Covid-19 Government Response
internal movement restrictions, international travel controls and public info Tracker, but refines the measure to account for can-
campaigns. The “plus” variant of the Stringency Index adds a tenth, i.e., facial
coverings, as well as refines workplace closures to incorporate reduced open- tonal differences within Switzerland. Figure 2 shows the
ing hours. For the detailed construction of the indices, we refer to Pleninger
et al. (2021) and the documentations provided there.
Goller and Wolter Swiss J Economics Statistics (2021) 157:6 Page 7 of 15
Table 3 Descriptive statistics
Mean SD Min Max
Outcomes
# Queries by day 25,386.61 11,513.79 3259 52,105
# Queries by day, canton and occupation 14.520 56.787 0 7353
Log(# Queries by day, canton and occupation)a 1.381 1.442 0 8.903
Covariates
Stringency Index 0.559 0.179 0.194 0.815
Stringency plus Index 0.554 0.167 0.175 0.750
Stringency over timeb 1.153 0.842 0.002 2.753
Public holidays 0.025 0.157 0 1
School vacation 0.256 0.437 0 1
Share signed apprenticeship contractsc 0.709 0.271 0.001 1
512,148 Observations. SD standard deviation
a
To avoid NA values due to undefined log(0) we chose to add + 1 to each value before log-transformation
b
Stringency Index multiplied by an in time increasing variable as described in the text
c
Due to missing information in the monitoring 389,844 observations
Stringency Index over the investigated time period exem- previous years total number as an additional control vari-
plary for the canton of Zurich. able. Descriptive statistics on some of these variables can
To investigate if the effect of the restrictions is (linearly) be found in Table 3.
declining or growing over time, a variable “Stringency In the following we analyse the relation between the
over time” is constructed as follows: shutdown and restriction variables and the number
of search queries in a linear regression framework. To
t
Stringencyovertime = Stringency × , check the dependence of the results on the choice of the
100 method in “Appendix”, the main results are replicated
where t = 1,…,T is an increasing numbered index for with the Stringency plus Index, as well as using Pseudo-
the respective day, i.e. is equal to 1 for the 28th February Poisson Maximum Likelihood (Santos Silva & Tenreyro,
2020, up to 402 for the 4th April 2021. Dividing by 100 2006).
leads to more interpretable coefficients, which can now
be seen as for every 100 days the Stringency Index coeffi- 5 Results
cient is changed by the Stringency over time coefficient.10 To investigate the influence of the restrictions imposed
To control for specific (COVID-19-unrelated) patterns, due to the COVID-19 pandemic on the supply of appren-
we construct several variables. Repeated weekly patterns tices, we analyse the correlation of the Stringency Index
are accounted for by including binary variables for each and the shutdowns on the number of daily search queries
day of the week, while seasonal variations are captured by canton and occupation. Important for the interpreta-
using monthly indicators. For general cantonal differ- tion of the coefficients is that the outcome, the number
ences canton dummies are included, as well as dummies of daily search queries, is used in a log-transformation.
to account for the occupation group in which the search Thus, for an increase of 1 unit in the Stringency Index we
was done. Since search intensity is generally lower on observe 61.7% less search queries in Panel A of Table 4,
public holidays, as well as school vacations those are column (1). Though, it is important to keep in mind that
also included on a cantonal level. To take into account an increase from 0 to 1 in the stringency would translate
that apprenticeship contracts are concluded at different to a switch from no restrictions to the maximum possi-
times in the year, depending on the region we include the ble restrictions (the maximum observed Stringency Index
already signed apprenticeship contracts as share of the value in the data was 0.815). To put this index into per-
spective, we have seen that the first shutdown in March
2020 resulted in an increase of about 0.5 units.11
10
The analysis of the time-dependency of the Stringency Index on outcomes
11
is motivated by the observation of a COVID-19 shutdown fatigue. Goldstein An increase in 0.5 is due to a bundle of imposed or tightened restric-
et al. (2021) e.g., found that there is a fading effect of the COVID-19 related tions. Some more concrete examples are: Closing all schools and universi-
restrictions on mobility and health outcomes, using a panel of about 150 ties increases the Stringency Index by 0.111, while allowing gatherings of 50
countries. instead of 10 people decreases the index by 0.028.
Goller and Wolter Swiss J Economics Statistics (2021) 157:6 Page 8 of 15
Table 4 Main results
(1) (2) (3) (4)
Panel A: Stringency Index
Stringency Index − 0.617*** − 0.683*** − 0.579*** − 0.569***
(0.014) (0.015) (0.017) (0.017)
Stringency over time 0.091***
(0.012)
Public holiday − 0.304*** − 0.303*** − 0.277*** − 0.280***
(0.009) (0.009) (0.009) (0.009)
School vacation − 0.208*** − 0.209*** − 0.200*** − 0.187***
(0.006) (0.006) (0.006) (0.006)
Share signed apprentice. contr − 0.604***
(0.059)
Observations 512,148 512,148 236,964 236,964
Time range 28.02.’20–04.04.’21 28.02.’20–31.08.’20
Panel B: Shutdowns
Stringency Index − 0.095*** − 0.191***
(0.019) (0.022)
Stringency over time 0.053**
(0.020)
Shutdown 1 − 0.394*** − 0.366*** − 0.378*** − 0.379***
(0.006) (0.007) (0.008) (0.006)
Shutdown 2 − 0.199***
(0.007)
Public holiday − 0.326*** − 0.324*** − 0.328*** − 0.277***
(0.009) (0.009) (0.009) (0.008)
School vacation − 0.206*** − 0.207*** − 0.208*** − 0.171***
(0.006) (0.006) (0.006) (0.006)
Observations 512,148 512,148 512,148 512,148
Time range 28.02.’20–04.04.’21
Outcome is the log (daily queries per canton and occupation). Other controls are dummies for: Days of the week, month, occupation, canton, year. Shutdown 1 is
defined as 16.03–11.05.2020. Shutdown 2 is defined as 22.12.2020–28.02.2021. Columns (3) and (4) in Panel A are restricted until apprenticeships 2020 started in the
beginning of September. Standard errors in parentheses are clustered standard errors on occupation x canton level. **, and *** marks statistical significance at the 1%,
and 0.1% level
Estimates for the whole sample in Panel A, column (1), Stringency Index measured restrictions on the daily
as well as column (3) for the time range until the appren- search queries is halved.
ticeships 2020 started in the beginning of September, are Although, the share of signed apprenticeship contracts
robust, with barely changing coefficients of − 0.617 and in the respective NUTS-2 region has the expected nega-
− 0.579, respectively. Putting these numbers into perspec- tive impact on search activities (Panel A of Table 4, col-
tive, for the first shutdown in March 2020 this translates umn (4)), there is almost no change in the effect of the
to a decrease of about 30% of search queries. This is in the restriction index on daily search queries. In other words,
range of the decrease in search queries during public holi- the restrictions themselves affected the search activities
days (with coefficients between − 0.277 and − 0.328), and of those still in search of an apprenticeship at the time
larger in magnitude compared to the effect of school vaca- the restrictions were imposed, in a similar magnitude
tions (with coefficients between − 0.171 and − 0.209). irrespective of the number of outstanding contracts at
To investigate the effect of the restrictions over time that time in their region.
(column (2) in Table 4, Panel A) includes the Stringency Panel B in Table 4 reports the conditional correlation
over time variable, showing that for each additional for the first shutdown (Shutdown 1) in column (1), which
100 days, the correlation between the Stringency Index can be interpreted as 39.4% less search queries during the
and the number of search queries becomes smaller in first shutdown. This is a larger decrease as suggested by
effect by 0.091. Therefore, after 1 year the effect of the the Stringency Index discussed before, not only capturing
Goller and Wolter Swiss J Economics Statistics (2021) 157:6 Page 9 of 15
Fig. 3 Treatment effect by occupation, 8 categories. Correlation of Stringency Index and daily queries; linear model—specification as in Table 4,
Panel A, column (1). Blue diamonds are the point estimates for the specific region accompanied by the 95% confidence interval. Vertical solid grey
line (with grey shaded confidence interval) is the population average effect. Eight categories according to the SwissDoc classification
the first shutdown, but changing throughout the full diminishes over time for other reasons, or if both played
observation period. Controlling for the first shutdown, in a role cannot be finally answered.
Panel B, column (2) leads to a substantial decrease of the Even though on a lower magnitude, column (3) in
correlation between the Stringency Index and the daily Table 4s’ Panel B we still find the decreasing correlation,
search queries compared to Panel A, column (1). Among with a positive Stringency over time coefficient (0.053).
the containment policies during the full time of the first For the second shutdown (Shutdown 2, Table 4, Panel
shutdown was the uniquely used closure of obligatory B in column (4)) we observe a decrease in search que-
schools. If this reduction in the influence of the equally ries of about 20%. This lower effect is also visible if we
weighted stringency measure on search queries was due investigate the shutdowns descriptively on a raw, daily
to the closures of obligatory schools, if the negative rela- basis in Figs. 7 and 8 in “Appendix 1”. We find a clear
tion between containment policies and search activity effect for Shutdown 1, regenerating slowly, while for
Goller and Wolter Swiss J Economics Statistics (2021) 157:6 Page 10 of 15
Shutdown 2 we observe a short and strong effect, which we are far from claiming causality for the relationship
is partly explainable by school vacations and the public between the search queries and the number of signed
holidays around Christmas and New Year. Defining dif- apprenticeship contracts per month, it is striking that
ferent lengths of the Shutdown 1, Fig. 6 in “Appendix” first the slowing and then the accelerating of the num-
shows that the effect was strongest two weeks after the ber of signed contracts between April and October 2020
shutdown on the 16th of March and steadily decreased coincides with the temporal pattern of search intensity
afterwards. just before these months on the national platform for
Tables 5 and 6 in “Appendix 2” investigate the sensitiv- apprenticeship vacancies. With regard to the develop-
ity of those results regarding the Stringency vs. Stringency ment of the Swiss apprenticeship market during the 2020
plus Index, and the estimation method using the PPML pandemic year, the analyses provide therefore an expla-
method. The results are comparable to our main results nation for the initial decline in apprenticeship contract
in Table 4 and the conclusions drawn do not change. signings at the beginning of the first shutdown and the
To investigate the robustness of the presented esti- subsequent catch-up effect.
mates, we provide subgroup estimates by types of occu- However, due to several limitations in the availability of
pations and NUTS-2 regions. Figure 3 investigates groups data and the uniqueness of certain important pandemic-
defined by eight occupation categories and shows differ- related restrictions, the findings are limited in at least
ent occupation types to react very similar to the restric- three respects. The answer to these three questions is
tions.12 Although effects differ by occupation categories, therefore the subject of future research.
they are all statistically significant, negative, and similar Firstly, we do not have similarly disaggregated data on
in magnitude. Figure 4 sheds some light on differential the demand for apprentices available at high temporal
effects by regions. The strongest effect is visible for the frequency for the period considered here. Even if we have
regions of Ticino and Région lémanique, which are also individual data points for this period that are sufficient as
the regions first and, at least in the beginning of the pan- an indication to assume that the demand for apprentices
demic, hardest affected by the Coronavirus. As discussed had not changed negatively after the first shutdown and
before, in these regions, apprenticeship contracts are also after the second shutdown, we cannot observe sup-
signed closer to the start of the apprenticeship. Results in ply and demand simultaneously.
Fig. 4 confirm this as the search intensity in these regions Secondly, we cannot separately examine the impact
is reduced most. Despite the results are heterogeneous, of school closures on search intensity because com-
it can be noted that all the regions show significant and pulsory schools were only closed once during the first
negative responses to the stringency of restrictions. shutdown and in all cantons simultaneously. Even
though it is very likely that the school closures had a
6 Conclusion and discussion particularly dampening influence on search intensity,
This paper examines the impact of the economic, social, and this would therefore also explain why later fluc-
and educational restrictions triggered by the COVID- tuations in the Stringency Index no longer had such a
19 pandemic on the behaviour of apprenticeship supply large influence on search intensity, other explanations
before, during and after these shutdowns. The behav- cannot be ruled out. On the one hand, and as with
iour of supply is analysed using a novel dataset formed other behaviours such as geographic mobility, it may
by around 10 million searches on the official public be that the same restriction measures have no longer
platform for apprenticeships. With this data, the search elicited the same responses from the population over
intensity for apprenticeships can be approximated and time, a phenomenon known as COVID-19-fatigue. In
tracked daily. The analysis shows a strong temporary our case, this could also be interpreted as meaning that
decline in search queries and therefore assumingly also the young people had recovered from an initial "shock"
of the supply of apprentices, during the first shutdown and knew afterwards how to deal with the crisis. On the
in mid-March 2020. The further development of the other hand, state support measures could also have had
search queries correlates strongly with the KOF Strin- an effect, by means of which school leavers in particular
gency Index, which shows the degree of restrictions were motivated to actively seek apprenticeships again
imposed daily and per canton. Furthermore, the analyses and in some cases also supported them in doing so.
also show that the influence of the restrictions on search This would also explain the increase in search activi-
intensity decreases substantially over time. Although ties that was already evident towards the end of the first
shutdown.
Thirdly, and beyond the scope of our analysis, the
question remains as to why the decline in GDP of
12
The eight occupation categories are chosen according to the super-catego- around three per cent in 2020, in contrast to much
ries in the official SwissDoc classification.
Goller and Wolter Swiss J Economics Statistics (2021) 157:6 Page 11 of 15
Appendix 1: Additional resources
See Figs. 5, 6, 7 and 8.
Fig. 4 Treatment effects by NUTS-2 regions. Correlation of Stringency
Index and daily queries; linear model—specification as in Table 4, Fig. 5 Share of signed apprenticeship contracts, April 2020. In April
Panel A, column (1). Blue diamonds are the point estimates for 2020 as a share of the number of previous years’ final number of
the specific region accompanied by the 95% confidence interval. signed apprenticeship contracts within a NUTS-2 region
Vertical solid grey line (with grey shaded confidence interval) is the
population average effect
smaller crises in the past, did not leave a trace on the
apprenticeship market. We can only speculate about
this at present. The only certainty is that rarely in eco-
nomic history has a recession started on a very specific
day and ended after a very specific period of time, as
decreed by the authorities, and at the same time accom-
panied by fiscal support measures that know no com-
parison in history. However, it can be deduced from
these correlations that the timing of the exogenous
shock (COVID-19 pandemic), namely at a time when
more than 60% of apprenticeship contracts for 2020
had already been signed, as well as the timing of the
transition to reopened schools contributed to the fact
that the economic slump in 2020 bypassed the appren-
ticeship market without leaving noticeable traces.
Finally, for the future monitoring of and research on
the apprenticeship market, it can be concluded from the Fig. 6 Shutdown 1 effects for different time periods. Treatment
analyses that the data on the search activities for appren- effects from separate regressions using different definitions, i.e.,
ticeship positions provide a very good high-frequency time spans of the treatment variable. Since (canton by occupation
clustered) standard errors are very small (< 0.006) they are omitted in
indicator to follow the development on the supply side of
this illustration
the apprenticeship market in "real time".
Goller and Wolter Swiss J Economics Statistics (2021) 157:6 Page 12 of 15
Fig. 7 First shutdown, daily queries, Feb 28–Jun 01, 2020. Green solid line connects for illustration purposes the daily queries (black points). Yellow
shaded area represents the time between 13.03 and 11.05.2020
Fig. 8 Second shutdown, daily queries, Oct 01, 2020–Mar 15, 2021. Green solid line connects for illustration purposes the daily queries (black
points). Yellow shaded area represents the time between 22.12.2020 and 28.02.2021
Goller and Wolter Swiss J Economics Statistics (2021) 157:6 Page 13 of 15
Appendix 2: Additional results
See Tables 5 and 6.
Table 5 Main results; Stringency plus
(1) (2) (3) (4)
Panel A: Stringency Index
Stringency plus − 0.664*** − 0.733*** − 0.620*** − 0.610***
(0.015) (0.016) (0.019) (0.018)
Stringency plus over time 0.094***
(0.012)
Public holiday − 0.304*** − 0.303*** − 0.279*** − 0.281***
(0.009) (0.009) (0.009) (0.009)
School vacation − 0.210*** − 0.210*** − 0.200*** − 0.187***
(0.006) (0.006) (0.006) (0.006)
Share signed apprentice. contr − 0.605***
(0.059)
Observations 512,148 512,148 236,964 236,964
Time range 28.02.’20–04.04.’21 28.02.’20–31.08.’20
Panel B: Shutdowns
Stringency plus − 0.086*** − 0.173***
(0.020) (0.024)
Stringency plus over time 0.063**
(0.021)
Shutdown 1 − 0.394*** − 0.371*** − 0.386*** − 0.379***
(0.006) (0.007) (0.008) (0.006)
Shutdown 2 − 0.199***
(0.007)
Public holiday − 0.326*** − 0.324*** − 0.329*** − 0.277***
(0.009) (0.009) (0.009) (0.008)
School vacation − 0.206*** − 0.207*** − 0.208*** − 0.171***
(0.006) (0.006) (0.006) (0.006)
Observations 512,148 512,148 512,148 512,148
Time range 28.02.’20–04.04.’21
Outcome is the log (daily queries per canton and occupation). Other controls are dummies for: Days of the week, month, occupation, canton, year. The Stringency over
time coefficient is estimated on a basis of 100 days. Shutdown 1 is defined as 16.03.–11.05. 2020. Shutdown 2 is defined as 22.12.2020–28.02.2021. Columns (3) and (4)
in Panel A are restricted until apprenticeships 2020 started in the beginning of September. Standard errors in parentheses are clustered standard errors on occupation
x canton level. ***Marks statistical significance at the 0.1% level
Goller and Wolter Swiss J Economics Statistics (2021) 157:6 Page 14 of 15
Table 6 Main results, PPML
(1) (2) (3) (4)
Panel A: Stringency Index
Stringency Index − 0.959*** − 0.917*** − 1.032*** − 1.031***
(0.022) (0.003) (0.029) (0.029)
Stringency over time 0.058**
(0.021)
Public holiday − 0.504*** − 0.506*** − 0.317* − 0.317*
(0.113) (0.113) (0.160) (0.161)
School vacation − 0.383*** − 0.382*** − 0.380*** − 0.376***
(0.009) (0.009) (0.016) (0.011)
Share signed apprentice. contr − 0.076
(0.205)
Observations 512,148 512,148 236,964 236,964
Time range 28.02.’20–04.04.’21 28.02.’20–31.08.’20
Panel B: Shutdowns
Stringency Index − 0.485*** − 0.467***
(0.028) (0.032)
Stringency over time 0.054***
(0.006)
Shutdown 1 − 0.497*** − 0.340*** − 0.367*** − 0.471***
(0.010) (0.013) (0.014) (0.010)
Shutdown 2 − 0.369***
(0.020)
Public holiday − 0.546*** − 0.530*** − 0.529*** − 0.461***
(0.113) (0.113) (0.113) (0.118)
School vacation − 0.385*** − 0.385*** − 0.386*** − 0.325***
(0.009) (0.009) (0.009) (0.010)
Observations 512,148 512,148 512,148 512,148
Time range 28.02.’20–04.04.’21
Outcome is the daily queries per canton and occupation. Other controls are dummies for: Days of the week, month, occupation, canton, year. The Stringency over time
coefficient is estimated on a basis of 100 days. Shutdown 1 is defined as 16.03.–11.05.2020. Shutdown 2 is defined as 22.12.2020–28.02.2021. Columns (3) and (4) in
Panel A are restricted until apprenticeships 2020 started in the beginning of September. *, **, and ***Marks statistical significance at the 5%, 1%, and 0.1% level
Abbreviations Declarations
VET: Vocational education and training; SDBB|CFSO: Swiss service centre for
vocational education and training; SERI: Swiss State Secretariat for Education, Competing interests
Research, and Innovation; COVID-19: New coronavirus; NUTS-2: Nomenclature The authors declare that they have no competing interests.
of Territorial Units for Statistics.
Author details
Acknowledgements 1
Centre for Research in Economics of Education, University of Bern, Bern,
We thank the SDBB|CSFO, especially Marc Fuhrer, for providing the data, valu- Switzerland. 2 Swiss Coordination Centre for Research in Education, Aarau,
able assistance, and support. Switzerland. 3 CESifo, Munich, Germany. 4 IZA, Bonn, Germany. 5 Swiss Insti-
tute for Empirical Economic Research, University of St. Gallen, St. Gallen,
Authors’ contributions Switzerland.
Both authors contributed equally to the idea, data analysis, results interpre-
tation and writing of the paper. Both authors read and approved the final Received: 3 May 2021 Accepted: 20 August 2021
manuscript.
Funding
This study was partly funded by the SERI through its "Leading House VPET-
ECON: A Research Center on the Economics of Education, Firm Behavior and References
Training Policies". Baker, S. R., Bloom, N., Davis, S. J., & Terry, S. (2020). Covid-induced economic
uncertainty. NBER Working Paper, w26983.
Availability of data and materials Balgova, M., Trenkle, S., Zimpelmann, C., & Pestel, N. (2021). Job search dur-
The data that supports the findings of this study are available from the ing a pandemic recession: Survey evidence from the Netherlands. IZA DP
SDBB|CFSO but restrictions apply to the availability of these data, which were No.14180.
used under license for the current study, and so are not publicly available. The Bolli, T., Oswald-Egg, M. E., & Rageth, L. (2021). Meet the need—The role of
KOF Stringency Index is available from the KOF High Frequency Dashboard vocational education and training for the youth labour market. Kyklos,
(https://kofdata.netlify.app/#/). 74(3), 321–348.
Goller and Wolter Swiss J Economics Statistics (2021) 157:6 Page 15 of 15
Brunello, G. (2009). The effect of economic downturns on apprenticeships and Luethi, S., & Wolter, S. C. (2020). Are apprenticeships business cycle proof? Swiss
initial workplace training: A review of the evidence. Empirical Research in Journal of Economics and Statistics, 156(3), 1–11.
Vocational Education and Training, 1(2), 145–171. Maier, T. (2020). Auswirkungen der “Corona-Krise” auf die duale Berufsausbildung:
Buechel, K., Legge, S., Pochon, V., & Wegmueller, P. (2020). Swiss trade during Risiken, Konsequenzen und Handlungsnotwendigkeiten. Bonn: Federal
the COVID-19 pandemic: An early appraisal. Swiss Journal of Economics Institute for Vocational Education and Training.
and Statistics, 156(22), 1–15. Marinescu, I., Skandalis, D., & Zhao, D. (2021). The impact of the federal pandemic
Burri, M., & Kaufmann, D. (2020). A daily fever curve for the Swiss economy. unemployment compensation on job search and vacancy creation. NBER
Swiss Journal of Economics and Statistics, 156(6), 1–11. working paper, No. w28567.
Chetty, R., Friedman, J. N., Hendren, N., Stepner, M., & The Opportunity Muehlemann, S., Pfeifer, H., & Wittek, B. (2020). The effect of business cycle
Insights Team. (2020). The economic impacts of COVID-19: Evidence from expectations on the German apprenticeship market: estimating the
a new public database built using private sector data. NBER working impact of Covid-19. Empirical Research in Vocational Education and Train-
paper, w27431. ing, 12(8), 1–30.
Coibion, O., Gorodnichenko, Y., & Weber, M. (2020). Labor markets during the Muehlemann, S., & Wolter, S. C. (2020). The economics of vocational training.
COVID-19 crisis: A preliminary view. NBER working paper, 27017. In S. Bradley & C. Green (Eds.), The economics of education (pp. 543–554).
Eckert, F., & Mikosch, H. (2020). Mobility and sales activity during the Corona Amsterdam: Elsevier.
crisis: Daily indicators for Switzerland. Swiss Journal of Economics and Muehlemann, S., & Wolter, S. C. (2021). Business cycles and apprenticeships.
Statistics, 156(1), 1–10. Oxford research encyclopedia of economics and finance. (forthcoming).
Forsythe, E., Kahn, L. B., Lange, F., & Wiczer, D. (2020). Labor demand in time Oeynhausen, S., Milde, B., Ulrich, J. G., Flemming, S., & Granath, R.-O. (2020). Die
of COVID-19: Evidence from vacancy postings and UI claims. Journal of Entwicklung des Ausbildungsmarktes im Jahr 2020. Bonn: Federal Institute
Public Economics, 189, 104238. for Vocational Education and Training.
Goldstein, P., Yeyati, E., & Sartorio, L. (2021). Lockdown fatigue: The diminishing Persson, J., Parie, J. F., & Feuerriegel, S. (2021). Monitoring the COVID-19 epidemic
effects of quarantines on the spread of COVID-19. Covid Economics, 67, with nationwide telecommunication data. arXiv:2101.02521
1–23. Pleninger, R., Streicher, S., & Sturm, J.-E. (2021). Do COVID-19 containment meas-
Goolsbee, A., & Syverson, C. (2021). Fear, lockdown, and diversion: Comparing ures work? Evidence from Switzerland. KOF Working Papers, No. 494.
drivers of pandemic economic deline 2020. Journal of Public Economics, Santos Silva, J., & Tenreyro, S. (2006). The log of gravity. Review of Economics and
193, 104311. Statistics, 88(4), 641–658.
Gupta, S., Montenovo, L., Nguyen, T. D., Lozano-Rojas, F., Schmutte, I. M., Simon, Wolter, S. C., & Ryan, P. (2011). Apprenticeship. In E. A. Hanushek, S. Machin,
K. I., Weinberg, B. A., & Wing, C. (2020). Effects of social distancing policy on & L. Woessmann (Eds.), Handbook of economics of education (Vol. 3, pp.
labor market outcomes. NBER working paper, 27280. 521–576). Amsterdam: Elsevier.
Kraenzlin, S., Meyer, C., & Nellen, T. (2020). COVID-19 and regional shifts in Swiss
retail payments. Swiss Journal of Economics and Statistics, 156(14), 1–20.
Lalive, R., Lehmann, T., & Siegenthaler, M. (2020). Die Schweizer Stellensuchenden Publisher’s Note
im COVID19-Lockdown. Working paper. https://e4s.center/wp-content/ Springer Nature remains neutral with regard to jurisdictional claims in pub-
uploads/2020/09/JobSeekersCOVID-full-DE.pdf lished maps and institutional affiliations.
File and source
- File
- too-shocked-to-search-the-covid-19-shutdowns-impact-on-the-search-for-apprenticeships.pdf
- Size
- 1,734,433 bytes
- SHA-256
- bdec3f993328257498a3ebefe71ed4ad0c2624789bf5e0ce5f7d5632694b9598
- Original
- doi.org