Experiences of return as mediated through the most popular news websites

Media analysis has been applied in migration research, including, for example, the refugee crisis (e.g. Berry, Garcia-Bianco, & Moore, 2015) and return migration to Lithuania (e.g. Budginaite, 2012b). This analysis has been carried out by coding the texts manually, using a pre-set coding structure, which requires a lot of resources and may even miss important topics. In this section we experiment with emerging big data-type tools and machine-based automatic text processing. Such tools allow for a more inductive approach; they can be applied quickly and on a large data-set. Admittedly, the data set used in our analysis (1017 articles dealing with return migration) is relatively limited. They key reason is that the sheer number of articles on return migration has not been very large, and some texts have already been archived, hidden, or deleted and are not available for scraping. Nevertheless, the data set is sufficient to explore the method as well as draw some inferences on questions of interest to this chapter.

Our data shows that the overall number of texts dealing with return migration has been on the increase (Figure 8.4). While specific numbers should be treated with caution, as older articles are more difficult to scrape, the trend itself is plausible. It is in line with the overall tendency of increasing return migration as well as higher interest in the topic from journalists and the society.

Number of articles on return migration scraped from the main news portals

Figure 8.4 Number of articles on return migration scraped from the main news portals

Notes. We could not assign a concrete date for 37 articles that most likely come from the period 2012-2015.

Source: Own elaboration.

Next, we have checked the overall sentiment of the media regarding the returnees. Word cloud analysis shows that overall rehirn migration received rather positive coverage. The dominating words are ‘good', ‘new', 'important', 'possible'. The negative words, such as 'difficult', ‘not be able to’, 'mistake', ‘suspected' are visible but less frequent. Analysis by portal and by time showed a certain variation. For example, the main business portal vz.lt had somewhat neutral coverage, featuring words such as 'service', ‘client', 'tax', 'construction'. The more positive stories featured words like 'new', ‘possible’, ‘important’, while the more negative reports included 'prohibited', ‘does not contribute'. There was some fluctuation in our dataset over the years. In 2017-2019 the key words were 'new’, 'now’, ‘good', 'true', ‘possible’, 'need', 'want', ‘return’, 'worked'. The words with clearly negative connotations were 'difficult', ‘bitter’, 'mistake', ‘suspected', 'not be able to’. We ran the word2vec algorithm specifically with the bi-gram 'acquired abroad'. It shows that words that can be used interchangeably with this bigram include ‘specific’, ‘ability’, 'professional', ‘patience’, 'deficiency’, 'diploma', 'international experience’.

We carried out topic modelling, based on the LDA algorithm, to see how different words in our texmal data cluster into topics. We firstly asked to generate one most prominent topic based on 20 words. The algorithm remrned a collection of words as presented in Table 8.4. While the meaning of the topic is subject to interpretation, the words seem to cluster into a narrative, saying that presently

Table 8.4 LD A topic modelling algorithm. 1 topic, 20 words

0.021*”is” + 0.007*”time” + 0.006*”person” + 0.006*”many” + 0.006*”in Lithuania"

+ 0.005*”emigrant” + 0.005*”have” + 0.005*’4vork” + 0.004*”country”

+ 0.004*”Lithuania" + 0.004*"may” + 0.003*”say” + 0.003*”good" + 0.003*”here”

+ 0.003 *”hello”

Source. Own elaboration

many persons who used to be emigrants have work in Lithuania and are satisfied with their return.

In order to carry out a more detailed analysis, we have varied the number of topics (up to 50) and changed the number of words per topic from 10 to 30. Based on this analysis, we singled out the following narratives that are most prominent in our dataset:

i Being in Lithuania, work, new life;

ii Bringing family and children back to Lithuania;

iii A returnee creating or having a business in Lithuania;

iv Health procedures and medical operations (emigrants returning for health checks or medical procedures);

v Owning property (a house, a flat) in Lithuania;

vi Emigrant suspected of a crime; car crash; investigation being carried out;

vii "Four million’ - a project by the national broadcasting company (LRT) about Lithuanian migrants

We also checked how various state and labour market instimtions (including the key policy instruments used by these institutions) feature in the texts on return migration. The assumption is that if an insthution is mentioned frequently and is presented in a positive context, then its role is visible and acknowledged as such in society. If this is not the case, it does not necessarily mean that an institution is not important; nevertheless, this is a useful proxy showing that an institution is not perceived as making much difference on the decisions and labour market status of the return migrants. In order to understand the context in which an institution is presented, we ran a word cloud analysis as well as the word2vec algorithm, which shows the words in our data set that are most similar to the selected institution.

As demonstrated in Table 8.5, central government institutions were mentioned most frequently in our dataset (Parliament, President, Government). It means that remrn migration, as well as the success or unsuccess of return, are associated frequently with the overall economic and social environment in the country. The words that feature most in our dataset alongside these institutions are: ‘new’, ‘important’, ‘good’, ‘big’, ‘service’, ‘need’, ‘tax’, ‘suggest’. These words are related to the modus operandi of the institutions (suggesting, making, or justifying decisions, responding to needs). Analysis based on the word2vec algorithm shows that, among other things, Parliament is discussed frequently in the context of the

Table S.5 The number of articles mentioning specific institution, programme, or concept

In Lithuanian"

No of articles

Central or municipal level institutions


Parlamentas. Seimas











Labour market institutions and organisations

Ministry of Social Affairs and Labour

Socialines apsaugos ir darbo ministerija, SADM


Public employment service

Darbo birza, Uzimtumo tarnyba


State Social Insurance Fund Board



Trade union

Profesine sajunga, profsajunga


Labour council

Darbo taryba


Labour market and related instruments

Social model

Socialinis modelis


Labour code

Darbo kodeksas


Collective agreement

Kolektyvinis susitarimas, kolektyvine sutartis


Unemployment benefit

Nedarbo pasalpa. nedarbo ismoka, bedarbio pasalpa


Minimum salary

Minimalus atlyginimas





Social insurance

Socialinis draudimas


Programmes aimed specifically at returnees

“I choose Lithuania”

„Renkuosi Lietuvq"


“Create for Lithuania”

Kurk Lietuvai


Important concepts





Svietimas, issilavinimas


Knowledge, skills

Zinios, jgudziai

















Acquired abroad

Uzsienyje jgytas


Low qualification, unqualified

Zema kvalifikacija. nekvalifikuotas


Source: Own elaboration.

Notes : * Given that the endings of words in Lithuanian change depending on the case, gender, and tense, we run the search using the root parts of these words.

demographic situation in the country. Govermnent is visible in the discussion on the economic aspects of migration and strategies to encourage return.

Further, we considered institutions and organisations that oversee implementing or mediating labour market policy. They include the Ministry of Social Affairs and Labour, State Social Insurance Board, the Public Employment Service, trade unions, and labour councils. Table 8.5 shows that some of these institutions featured frequently in the texts on return migration, especially the PES. The word clouds concerning the PES and the Social Insurance Fund Board are very illustrative, with the following words dominating: 'need', ‘pay’, ‘contribution payment’, 'debt’, 'now’. The trade unions and labour councils also featured quite frequently in the dataset. It is an unexpected finding given the low trade union membership in the country; further, the majority of labour councils have been established only since 2017, when they became mandatory for companies with 20+ employees.12 The word clouds regarding these organisations are quite telling in that the dominating words are 'wish', 'need', 'not be able to’, or 'be impossible'. They seem to point to a tension between the potentially helpful role of these organisations as well as the underlying difficulties.

Finally, we checked various policy instruments, such as pensions, social insurance, minimum salary, unemployment benefit, training, and others, hi this paragraph we present only the more interesting or more conclusive evidence. Thus, pensions were mentioned frequently in articles on return migration. The word cloud and word2vec analysis showed that pensions are considered small and compared frequently to the minimum salary or social benefits. The word cloud on articles featuring the unemployment benefit has linked it to the words 'need', 'want', 'work', while the word2vec algorithm revealed that 'very small’ is used almost interchangeably with 'unemployment benefit'. The programme 'I Choose Lithuania', which was created to facilitate return, has also been noticed. It is associated with the words such as ‘want’, 'need', ‘returning’, ‘now’, 'possible’, 'important', 'difficult' and is seen through the narrative of offering the returnees help to overcome difficulties.

We also checked to what extent and in which capacity important concepts such as ‘entrepreneurship’, 'business', ‘work’, and 'unemployment' appear in our data set. The word ‘work’ is featured in almost all articles, showing that finding, having, and doing work is considered very important when discussing returnees. The word2vec algorithm shows that work is often used interchangeably with 'well paid', 'will create it ourselves’, as well as words such as 'need', 'search', ‘overtime’. The word 'business' goes often together with the words 'returned emigrants'. Words such as 'low qualification', 'unqualified' are used often in the texts. However, interestingly, the closely associated words in the word2vec analysis were 'need’, 'lack employees', ‘difficult find'.

Further, analysis shows that the concepts of knowledge, skills, and productivity appear often in the articles and are associated frequently with ‘new’, 'important', 'create', ‘offer’, 'need'; yet also 'do not contribute' was quite visible in the texts. The word2vec analysis shows that 'knowledge' can be used quite interchangeably with 'work offer’, while 'skills' is used closely with phrases like 'professional', 'quality', 'gained abroad', 'well paid'. When it comes to productivity, it is often used together with words such as 'new', 'need', ‘important’; 'high’, 'increasing'. Anumber of texts, however, point towards 'smaller' or 'decreasing' productivity as well as 'loss making', although this may reference the texts on increasing salaries, the middle-income trap, and the need to increase productivity to stay competitive.

< Prev   CONTENTS   Source   Next >