The Role of Social Media Content in Migration Aspirations: Mixed-Methods Evidence From Two Senegalese Regions

Interview Data

We conducted semi-structured interviews in the Dakar and Casamance regions, Italy (Turin, Rome, and Palermo), and Germany (Berlin and online due to COVID-19 regulations). The interviews were conducted between October 2019 and February 2020 in Senegal, between September 2020 and August 2021 in Germany and between September and October 2021 in Italy. The interview languages were French, Italian, Pulaar (a local Senegalese language for which an interpreter was employed), or Wolof (Senegal's lingua franca, spoken by one of the co-authors). In Senegal, 41 interviews were conducted with 57 individuals, amongst them migrants’ family members and friends (12 interviews), return migrants (24 interviews), and experts (5 interviews). In Italy, interviews were conducted with 28 individuals (26 Senegalese migrants and two migration experts). Finally, in Germany, 15 interviews were conducted with 19 individuals (5 Senegalese migrants and 14 members of Senegalese migrant associations).

At this point in time, the ethics committee of our institution does not require formal ethical clearance for qualitative research projects. Nevertheless, we discussed the interview guide with Senegalese researchers and adjusted it according to their feedback. Prior to the interview, the interview partners were informed, in written and oral form, about the scope of the research project, the anonymization procedure, and the possibility to interrupt or withdraw from the interview at any time. The interview partners gave their written or, in case of illiteracy, oral consent before the start of the interview. Inter alia, we asked interviewees about their (retrospective) migration aspirations, expectations, and experiences. During the qualitative fieldwork, social media came up frequently, so we included questions on interviewees’ exposure to and perception of social media content produced by Senegalese in Europe.

Interview partners were recruited through personal contacts and through gatekeepers such as migrant associations and a Senegalese research assistant. In some cases, interviewees also connected the interviewer with other potential interview partners (snowball sampling). In general, the sampling of interview partners in Senegal was guided by the desire for equal representation of potential migrants and return migrants, as well as of interviewees from Dakar and Casamance. Finally, we also aimed for equal representation along gender and educational lines when selecting potential interview candidates in all three countries. The interviews were transcribed and analyzed using MAXQDA, a software tool designed for computer-assisted text analysis. We coded the first five interviews inductively to develop a code tree for the analysis of the other interviews.

Survey Data

Between November 2021 and January 2022, survey data were collected in Dakar and the Casamance regions, the former being a very urban and the latter a rural region; our results are thus only representative for these parts of Senegal. Both regions have high emigration rates but differ greatly with regards to, e.g., poverty rates, education level, as well as ethnic and religious composition. The survey interviews were conducted in French or a local language (Wolof, Pulaar, Mandinka, or Diola) as computer-assisted personal interviews by local interviewers in cooperation with Senegal's National Agency of Statistics and Demography (ANSD). Respondents were randomly drawn from the population of 15- to 35-year-olds. Young respondents are expected to be especially technophilic, making it more plausible that social media shapes their migration aspirations. What is more, young respondents are most likely to migrate in the first place (e.g., Aslany et al. 2021). According to the World Bank (2024), in 2020, approximately 35 percent of the population was between 15 and 35, while only 25 percent were older than 35; approximately 40 percent were younger than 15. While our findings do not generalize to the whole adult population, especially people older than 35, 15 to 35 years old still account for approximately 71 percent of the adult population.

4,037 respondents from 4,037 households in the Dakar region (N = 997) and Casamance region (total N = 3,040, with the sub-regions Ziguinchor N = 1,629, Sédhiou N = 614 and Kolda N = 797) were surveyed. ² Within households, respondents were selected through a simple random draw, with the selected respondent having a two-thirds probability of being male. For all surveyed regions, ANSD used auxiliary data to validate that the survey data were indeed representative of the population in the surveyed regions and in the given age bracket (after accounting for the previously mentioned gender imbalance due to our survey design). In light of oversampling of men and random stratified sampling in several Senegalese regions, for all subsequent empirical analyses with the survey data, we employ post-stratification weights to reflect the actual gender distribution in Senegal (which is approximately 1:1), while also accounting for differences in population size (per the most recent census) between the surveyed regions. For instance, this results in female respondents from the Dakar region being weighted especially strongly.

In the survey, respondents were asked about their migration aspirations and plans as well as their socio-economic and demographic characteristics (e.g., age, gender and relationship status). Further survey questions concerned, for example, the respondents’ experience with migration as well as their thoughts on life in Senegal and abroad (e.g., with respect to economic prospects at home vis-à-vis abroad).

We accounted for ethical considerations when designing our survey in several ways. For one, we received formal ethical clearance for our research project and underlying questionnaire from the ethics committee of our institution before the start of the survey. For another, we were aware of the fact that especially inquiring about migration aspirations can be sensitive in a context in which many people cannot act on their migration aspirations. Consequently, we solicited feedback from ANSD and other Senegalese researchers prior to the data collection regarding the locals’ understanding of the questionnaire. This feedback proved very helpful, allowing us to change questions that were regarded as potentially sensitive to avoid potential discomfort for the survey participants. Finally, when the survey data were collected, all survey respondents were asked for their consent before the interviews started, where anonymity was always guaranteed.

Migration Aspirations

Migration aspirations are prevalent in Dakar and the Casamance. During the qualitative fieldwork, many interview partners declared their aspiration to migrate. Returned migrants spoke about their previous migration aspirations before leaving, but also often communicated that they wanted to migrate again in the future. Furthermore, parents spoke about their children's wishes to migrate, students about their plans to study abroad and many others about their own or their friends’ aspirations to migrate. As one interview partner puts it: “It is certain, these young people want to migrate” (Abdou, potential migrant, Casamance, 5 October 2019). ³

Migration aspirations might also be an expression of the imagined possibility of traveling and potentially be a consolation to the fact that many have little chance of actually migrating to Europe. Nevertheless, these migration aspirations are sometimes strong enough to make people pursue their migration goals despite many obstacles. For instance, Khady, a young woman, explained how her peers engage in various attempts to migrate irregularly to Europe:

There are many young people, they have arrived for example to the barrier of Spain or Morocco, they cannot continue, yet they return. They come back. But they try another time. They say to themselves ‘No, I have to get there, I have to get there’. They would rather die than give up. (Khady, potential migrant, Casamance, 5 October 2019)

Figure 1 provides a first look at our survey data concerning migration aspirations, where we create five bins that each correspond to a twenty-unit interval on the initial answer scale. We find that approximately 30 percent of respondents have no to little aspiration to migrate (0–40 on the initial scale), while approximately 50 percent have very strong aspirations to do so (>80). These numbers are comparable to those of the seventh wave (2016–2018) of the Afrobarometer, where Senegalese respondents were asked about their thoughts about migration to another country, with approximately 40 percent of respondents stating that they have considered moving to another country to live there. ⁵ Comparing Dakar and the Casamance, migration aspirations are more pronounced in Dakar (see the Supplementary Material, Appendix A).

OPEN IN VIEWER

Figure 1.

Migration Aspirations Among Survey Respondents.

Exposure to Migrants’ Social Media Content

In Dakar and the Casamance, many people use the internet and accordingly, also browse social media content. As Bineta, an interview partner in the capital Dakar, explains, social media is also accessible in rural areas such as the Casamance region, where we conducted part of the quantitative fieldwork:

Because there are people even who live in a village now that are connected. They say there's a high penetration rate of internet. It's not an internet rate but it's a strong penetration rate of Facebook and WhatsApp and Instagram and Snapchat. People use that mainly to chat, to exchange, to see what everyone else is doing every day. Those who are in Europe, they spend their time doing Snapchat and posts. (Bineta, potential migrant, Dakar, 19 November 2019)

In our quantitative survey, we similarly find that exposure to social media content produced by Senegalese migrants in Europe is fairly common. We measure this exposure by respondents’ answer to the following survey question: “How often do you see photos or videos by Senegalese living in Europe on social media? With social media, we mean Facebook, WhatsApp, Instagram, TikTok, Snapchat, and similar apps or websites.” The answer ranges from 0 (“never/practically never”) to 6 (“more than once a week”). Answer options in between are 1 (“less often than a year”), 2 (“once a year”), 3 (“once every three to six months”), 4 (“once a month”), and 5 (“once a week”). As shown in Figure 2, while about 50 percent of respondents never or rarely view social media content produced by Senegalese living in Europe, about 20 percent of respondents view it at least once a week. Comparing Dakar and the Casamance, respondents in Dakar tend to view such content more often than their counterparts from the Casamance; in particular, respondents in Dakar are much less likely to have (practically) never seen social media content by Senegalese migrants living in Europe (see the Supplementary Material, Appendix A).

OPEN IN VIEWER

Figure 2.

Social Media Use of Survey Respondents.

Qualitative Results

In our qualitative fieldwork, we find that both aspiring and returned migrants name social media content as an important factor fostering or having fostered their migration aspirations. For example, Seydou, the father of an aspiring migrant explains the origin of his son's motivation to migrate as follows: “They [his son's friends in Europe] would post pictures and they would impress him from those pictures, and it is by seeing those pictures that he is considering going” (Seydou, returnee, Casamance, 3 October 2019). This suggests that photos posted by the son's friends in Europe may have allowed his son to cognitively migrate and to imagine his potential life abroad, which becomes a driver for his aspirations. Similarly, Ibrahima, a returned migrant from Casamance, told us: Interviewer:

And how does *name of contact in Germany* talk about Germany? Does he send you pictures?

Furthermore, Senegalese migrants currently living in Europe recall that the social media content of other migrants indeed influenced their migration decision-making. For instance, Fatima, a Senegalese woman living in Italy, describes how pictures she saw on Facebook when she was still living in Senegal influenced her decision to migrate. According to her, social media content and especially photos play a key role in the formation of migration aspirations for people in Senegal:

To tell you the truth when I was in Senegal sometimes a person takes beautiful pictures and puts them in Facebook which will disturb your head, you will only want to leave, that's what you will want, yes, you only want to leave. (Fatima, migrant, Rome, 22 September 2021)

Quantitative Regression Analysis

Empirical Model

Next, we analyze the survey data by means of a regression analysis to assess whether individual migration aspirations can be systematically explained by the viewing of social media content of Senegalese migrants in Europe. ⁶ We estimate the following model:

m i g r a t i o n i = β 1 * s o c i a l m e d i a i + β ′ * X i ′ + ε i

(1)

Here, migration refers to the respondent's answer to the question on how much they have considered moving to another country, with the answer being on a scale from zero to 100 (higher values mean stronger migration aspirations). Our main variable of interest, socialmedia, refers to the respondent's answer regarding their use of social media to view content produced by Senegalese migrants in Europe (higher values coincide with higher levels of usage).

Additionally, our empirical model includes a vector of controls, X. The choice of these variables follows from a review of the determinants of migration aspirations by Aslany et al. (2021). All control variables come from our original survey conducted in Dakar and the Casamance. We control for respondents’ age, gender, and relationship status, where we expect age, female gender, and being in a long-term relationship or marriage to correlate with lower migration aspirations (Aslany et al. 2021). We also control for education with a dummy variable that is equal to unity when a respondent has secondary education or higher. We anticipate higher education levels to correlate with stronger migration aspirations (Aslany et al. 2021). Finally, we consider the influence of respondents’ expectations about the future of the Senegalese economy, using respondents’ thoughts about the economic situation in Senegal now vis-à-vis in three to five years (higher values correspond to a more positive economic outlook). Individuals who have more optimistic expectations are expected to have lower migration aspirations (Aslany et al. 2021).

Main Regression Results

As shown in Table 1, we find that viewing social media content by Senegalese migrants in Europe correlates with higher migration aspirations of individuals surveyed in Dakar and the Casamance. Our baseline estimates reported in Model (1) imply that a one-unit increase in the social media variable is associated with an increase in individual migration aspirations by approximately 2.4 points. By comparison, a one-unit (i.e., one-year) increase in age is linked to a decrease in individual migration aspirations by about 0.9 points.

The findings regarding the controls are as expected. We find that migration aspirations are higher among young, male, and educated individuals who are not in a relationship. As a first robustness check, in the Supplementary Material (Appendix B), we show that the positive relationship between social media usage and migration aspirations is not affected by the inclusion of further controls, indicating, e.g., personal life satisfaction, employment situation, trust in the Senegalese government, department-fixed or interviewer-fixed effects. ⁷

OPEN IN VIEWER

Table 1.

Main Regression Results.

	(1)	(2)	(3)	(4)	(5)
Sample →	Pooled	Pooled	Dakar	Casamance
Social media use	2.424***	2.218***	2.184***	2.598***	[0.44]
	(0.362)	(0.370)	(0.527)	(0.336)	[0.51]
Dakar		3.495
		(2.765)
Social Media Use × Dakar		0.011
		(0.637)
Age	−0.874***	−0.828***	−0.771***	−0.929***	[0.28]
	(0.167)	(0.169)	(0.255)	(0.155)	[0.60]
Female	−9.938***	−10.330***	−12.205***	−4.593***	[6.61]
	(1.706)	(1.663)	(2.505)	(1.591)	[0.01]**
Secondary education	6.485***	6.134***	4.783*	7.275***	[0.56]
	(1.972)	(1.993)	(2.834)	(1.776)	[0.46]
Economic expectations	−0.763	−0.596	0.052	−2.810***	[4.22]
	(0.851)	(0.864)	(1.155)	(0.784)	[0.04]**
In relationship	−10.335***	−9.892***	−10.628***	−8.509***	[0.28]
	(2.285)	(2.312)	(3.443)	(2.012)	[0.59]
Observations	4,037	4,037	997	3,040
R 2	0.129	0.130	0.113	0.101

Notes: OLS regression results reported. Dependent variable is the respondent's answer to the question “How much are you considering moving to another country to live there?” Responses to this question are on a scale from 0 to 100. Social media use variable measures respondents’ social media exposure by their answer to the following survey question: “How often do you see photos or videos by Senegalese living in Europe on social media? With social media we mean Facebook, WhatsApp, Instagram, TikTok, Snapchat and similar apps or websites.” The answer range is from 0 (“Never/Practically never”) to 6 (“More than once a week”). Answer options in between are 1 (“Less often than a year”), 2 (“Once a year”), 3 (“Once every three to six months”), 4 (“Once a month”) and 5 (“Once a week”). In (5), we report χ²-statistics and p-value associated with comparing the regression coefficients of Models (3) and (4), testing for differences between the urban and rural samples that we usually pool in our empirical analysis; an insignificant χ² indicates that the differences between both samples with respect to a specific explanatory variable are not statistically significant. Constant not reported. Heteroskedasticity-robust standard errors in parentheses. ***p < .01, **p < .05, *p < .1.

As another robustness check, we consider whether the use of different regression weights matters to our empirical results. For instance, due to relative population and sample sizes, observations by respondents from Dakar (especially women) are weighted more strongly than observations from the Casamance. This could easily yield extreme weights, potentially affecting our estimates. Reassuringly, in the Supplementary Material (Appendix B), we show that various methods of trimming extreme weights (e.g., by replacing extreme weights in the top 10 percent of observations with the trimmed maximum, i.e., the 90th percentile) produce empirical results reported almost identical to those reported in Table 1. Another way to probe the robustness of our findings is to apply weights so that observations from Dakar and Casamance each have the same weight. Here, the idea is that our estimates should be shaped equally by the urban (Dakar) and rural (Casamance) parts of our sample. ⁸ Again, we can show in the Supplementary Material (Appendix B) that using this weighting method produces empirical findings that are very similar to those reported in Table 1.

At the same time, our discussion about weights for the urban and rural parts of our sample also points to the issue of pooling data from Dakar and the Casamance. On statistical grounds, pooling increases sample size, which, in turn, increases the precision of our estimates (lower standard errors). However, the urban and rural regions differ—as shown above—with respect to their migration aspirations, social media use as well as other socio-economic and demographic characteristics. Consequently, pooling may come at the potential cost of masking meaningful differences between urban and rural regions with respect to the social media-migration aspiration nexus. For instance, the impact of social media use on aspirations might be smaller in Dakar, where social media use is more common and where other transnational influences might also shape migration aspirations.

One way to test this assumption is to interact with our social media use variable for Dakar. As shown in Model (2) in Table 1, there is no evidence of an interaction effect that would point to differences between the urban and rural regions with respect to social media use. Furthermore, in Models (3) and (4), we run separate regressions for the Dakar and Casamance sub-sample, comparing effect sizes between both models in a seemingly unrelated regression framework. While there are differences in effect sizes for some control variables (female gender and economic expectations), for our main explanatory variable of interest—social media use—effect sizes between both sub-samples are statistically indistinguishable from each other, while this variable also remains a statistically significant and positive predictor of migration aspirations for both sub-samples. Again, this suggests that viewing social media by Senegalese migrants in Europe is positively and in a similar manner associated with migration aspirations in Dakar and the Casamance. ⁹

Measurement of Social Media Use

In the following, we further probe the robustness of our findings. Here, one may not consider it appropriate to treat social media use as a linear variable, as we did in Table 1. Thus, as a robustness check, we consider alternative operationalizations of our main explanatory variable. We create a binary variable to differentiate between those who have never seen photos or videos by Senegalese living in Europe on social media and those who have. Moreover, we create a categorical variable that differentiates between individuals that view this type of media at most once a year, month or week, respectively, with the baseline being individuals that view it less than once per year. Additionally, we construct a variable that distinguishes between occasional users (Europe-specific content is at least viewed once a month, but not weekly), regular users (who browse this type of content at least once a week) and irregular users (who view social media content by Senegalese migrants in Europe at a maximum every 3–6 months or less). Finally, we compare respondents who browse social media content by Senegalese migrants residing in Europe at least once a week (heavy users) to those who do not.

In Table 2, we show that regardless of how we operationalize our explanatory variable, higher levels of use of migrants’ social media content correlate with higher migration aspirations. For instance, being a heavy user is associated with an increase in individual migration aspirations by approximately 11.5 points in comparison to the group of non-heavy users.

OPEN IN VIEWER

Table 2.

Viewing of Migrants’ Social Media Content and Migration Aspirations.

	(1)	(2)	(3)	(4)	(5)	(6)
Social media use						2.175***
						(0.380)
Social media user	15.111***
	(1.978)
Use at most once a year		13.657***
		(2.694)
Use at most once a month		10.359***
		(3.132)
Use at least once a week		16.918***		11.459***
		(2.191)		(1.875)
Occasional user			2.642
			(3.438)
Regular user			11.804***
			(1.941)
Social media and few					1.858***
Contacts to Europe					(0.574)
Social media and many					2.561***
Contacts to Europe					(0.377)
[F-Test]		[2.63]	[6.92]		[1.64]
[p-value]		[0.07]*	[0.00]***		[0.20]
Age	−0.841***	−0.858***	−0.899***	−0.904***	−0.886***	−0.908***
	(0.166)	(0.165)	(0.168)	(0.167)	(0.167)	(0.174)
Female	−10.036***	−10.016***	−10.166***	−10.212***	−9.935***	−9.671***
	(1.699)	(1.695)	(1.709)	(1.710)	(1.706)	(1.761)
At least secondary education	5.976***	5.750***	6.868***	6.894***	6.279***	7.461***
	(1.968)	(1.962)	(1.965)	(1.961)	(1.967)	(2.064)
Economic expectations	−0.967	−0.900	−0.778	−0.771	−0.714	−0.344
	(0.844)	(0.849)	(0.853)	(0.854)	(0.855)	(0.900)
In relationship	−10.235***	−10.110***	−10.544***	−10.601***	−10.283***	−9.718***
	(2.268)	(2.265)	(2.287)	(2.284)	(2.289)	(2.395)
Difference Dakar-Casamance χ2	0.70	1.31	1.43	0.47	0.66	0.92
(p-value)	(0.40)	(0.52)	(0.49)	(0.49)	(0.72)	(0.34)
Observations	4,037	4,037	4,037	4,037	4,037	3,713
R 2	0.136	0.139	0.125	0.125	0.130	0.126

Notes: OLS regression results reported. See Table 1 for operationalization of migration aspiration variable. See the main text for discussion of operationalization of social media use variables. F-statistic refers to a test of equality of coefficients (rejection of null hypothesis implies that coefficients are statistically different). Model (6) excludes from sample respondents that lived abroad in the past. The reported χ²-statistic is associated with testing whether the effect size of social media use on migration aspirations differs between Dakar and Casamance; an insignificant test statistic implies that effect sizes are indistinguishable between the urban and rural sub-sample in a statistical sense. To carry out this test, we run additional regressions of the models above for the Dakar or Casamance sub-sample, respectively. These regression results are reported in the Supplementary Material (Appendix B). Constant not reported. Heteroskedasticity-robust standard errors in parentheses. ***p < .01, **p < .05, *p < .1.

In Table 2, we also report χ²-statistics associated with testing whether the effect size of social media use on migration aspirations differs between the Dakar and Casamance sub-samples; an insignificant test statistic implies that effect sizes are indistinguishable between the urban and rural sub-sample in a statistical sense. To conduct the tests for differences in coefficients, we run additional regressions—as in Models (3) and (4) of Table 1—for the Dakar or Casamance sub-sample. For the sake of brevity, we do not report the results for these sub-sample regressions in the main text but report them in the Supplementary Material (Appendix B). Reassuringly, the associated tests always suggest that there are no differences in effect sizes between the Dakar and Casamance sub-samples when using alternative operationalizations of social media use.

Next, one may argue that social media content has a differential effect on aspirations depending on whether respondents have direct contact with Senegalese living in Europe. To study this argument, we consider the respondents’ answer to the survey question “How often do you communicate with Senegalese living in Europe?,” categorizing those who have such communication at least once a week as having many direct contacts and those who have such communication at most once a month as having few direct contacts. In Model (5), we show that both respondents with many and few contacts with Senegalese in Europe see stronger migration aspirations due to their viewing of specific social media content produced by Senegalese in Europe, with the difference between the two groups not being statistically significant. ¹⁰ This finding also holds for sub-samples that only consider data for Dakar or the Casamance, respectively (see Appendix B in the Supplementary Material for the full regression results).

Finally, one may argue that individuals who have lived abroad in the past (return migrants) may be less susceptible to the “siren call” of social media; including them in our sample may thus affect our estimates. To examine this argument, we consider the following question: “Have you ever lived elsewhere than here?,” categorizing those who named another country as having lived abroad. In Model (6), we show that dropping those respondents with foreign experience from the sample does not affect our main empirical conclusion. ¹¹ In addition to the pooled sample, this finding also holds for the Dakar and Casamance sub-samples (see Appendix B in the Supplementary Material for the full regression results).

Alternative Dependent Variables. Having established that different operationalizations of social media use yield similar results—using the pooled and sub-sample data—we next consider whether the positive association between the browsing of migrants’ social media content and individual migration aspirations may be due to the choice of the dependent variable. Here, we convert our original dependent variable (which measures aspirations on a scale from 0 to 100) into (1) a five-binned-variable, where each bin corresponds to a twenty-unit interval on the initial answer scale, (2) two binary variables that are equal to unity when a respondent has at least considerable migration aspirations (meaning a value of >60 on the initial answer scale) or has very strong aspirations (meaning a value of >90 on the initial answer scale), respectively, and (4) a variable that categorizes our initial dependent variable by its quartiles.

Moreover, there is a discussion in the literature on how to measure migration aspirations in the first place (see, e.g., Mjelva and Carling 2023). For instance, empirical operationalizations of migration aspirations may differ with respect to the nature of the respondent mindset, inquiring, e.g., considerations (as we do in our study), expectations, willingness, or planning (Mjelva and Carling 2023), which could lead to potentially different conclusions on the role of migrants’ social media content in individual migration aspirations. To address this concern, we employ three additional migration variables. First, we use the respondents’ answer to the following question as an alternative dependent variable: “Have you made concrete plans to move to another country within the next 12 months?” The answer is a binary variable, with unity indicating that the respondent has made concrete plans, while zero indicates that they have not. Second, we consider the question “Have you made any preparations to move to another country within the next 12 months?” Here, we create a binary variable that is equal to unity when a respondent has made preparations. Planning and preparations refer to different migrant mindsets, e.g., reflecting a respondent's desire to act on their migration aspirations (e.g., Tjaden, Auer, and Laczko 2019). Furthermore, we create a binary variable that is equal to unity when a respondent has strong migration aspirations (meaning a value of >90 on the initial answer scale) and concrete migration plans. This is to combine survey information on the consideration and planning mindsets.

We report our empirical findings using our alternative outcome variables in Table 3; depending on the nature of the dependent variable, we run logistic or ordered logistic regressions. Using pooled data, we find that viewing social media content of Senegalese migrants in Europe positively correlates with individual migration aspirations regardless of operationalization. We also find that respondents who browse social media by Senegalese in Europe are more likely to have made concrete plans or migration preparations to move to another country within the next 12 months. Finally, these respondents are also more likely to report both strong migration aspirations and concrete travel plans at the same time. As above, we report χ²-statistics associated with testing whether effect sizes of social media use on migration aspirations differ between the Dakar and Casamance sub-samples, finding that this is never the case (see Appendix B in the Supplementary Material for the full regression results for the two sub-samples). In summary, these findings suggest that our main empirical conclusion from Table 1 is robust to alternative operationalizations of the outcome variable.

OPEN IN VIEWER

Table 3.

Different Operationalizations of Outcome Variable.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Dependent Variable →	Aspirations 0–4	Aspirations 0–1	High aspirations 0–1	Quartile aspirations 0–3	Migration plan 0–1	Migration preparation 0–1	Aspirations and plans 0–1
Social media use	0.111***	0.108***	0.090***	0.116***	0.158***	0.195***	0.169***
	(0.019)	(0.021)	(0.020)	(0.019)	(0.024)	(0.029)	(0.026)
Age	−0.042***	−0.041***	−0.039***	−0.043***	0.000	0.002	−0.006
	(0.009)	(0.010)	(0.009)	(0.009)	(0.012)	(0.014)	(0.013)
Female	−0.516***	−0.522***	−0.446***	−0.504***	−0.652***	−0.600***	−0.738***
	(0.089)	(0.097)	(0.096)	(0.090)	(0.128)	(0.154)	(0.141)
Secondary education	0.290***	0.281***	0.199*	0.278***	0.471***	0.346**	0.380***
	(0.100)	(0.108)	(0.106)	(0.101)	(0.135)	(0.157)	(0.144)
Economic	−0.038	−0.059	−0.029	−0.042	−0.111*	−0.192***	−0.136**
Expectations	(0.045)	(0.047)	(0.045)	(0.046)	(0.057)	(0.068)	(0.058)
In relationship	−0.505***	−0.601***	−0.472***	−0.445***	−0.408**	−0.220	−0.528***
	(0.114)	(0.120)	(0.120)	(0.115)	(0.169)	(0.200)	(0.189)
Difference Dakar-Casamance χ2	0.60	0.07	1.96	0.90	0.15	0.11	0.04
(p-value)	(0.44)	(0.79)	(0.16)	(0.34)	(0.70)	(0.74)	(0.84)
Estimation method	Ordered Logit	Logit	Logit	Ordered Logit	Logit	Logit	Logit
Observations	4,037	4,037	4,037	4,037	4,037	4,037	4,037
Pseudo-R2	0.048	0.081	0.060	0.056	0.075	0.072	0.086

Notes: See main text for discussion of operationalization of dependent variables. See Table 1 for operationalization of social media use variable. The reported χ²-statistic is associated with testing whether the effect size of social media use on migration aspirations differs between Dakar and Casamance; an insignificant test statistic implies that effect sizes are indistinguishable between the urban and rural sub-sample in a statistical sense. To carry out this test, we run additional regressions of the models above for the Dakar or Casamance sub-sample, respectively. These regression results are reported in the Supplementary Material (Appendix B). Constant not reported. Heteroskedasticity-robust standard errors in parentheses. ***p < .01, **p < .05, *p < .1.

Exploration of Mechanisms from Interview Data

Pictures and videos of their journey posted by migrants in Europe on social media can help potential migrants to envision a specific (safe) travel route that others have already successfully used. As Seydou, a returned migrant, explains:

In general, when I started, I took the trouble to ask [using my phone] those who have already arrived at the destination. They are the ones who tell me where to go to get there. They are the ones who give me the route to follow. (Seydou, returnee, Casamance, 3 October 2019)

Most interview partners in Dakar and the Casamance describe the photos posted by migrants in Europe as beautiful or otherwise pleasant. For instance, Bineta describes the social media behavior of migrants as follows:

[…] the social networks they use is Facebook, WhatsApp, is Instagram. And we know in advance that on these networks, we will show only the pretty pictures, only positive posts. (Bineta, potential migrant, Dakar, 19 November 2019)

Indeed, most interviewees describe the social media content of migrants in Europe as positive. People in Dakar and the Casamance mostly only see posts, pictures, and videos that highlight positive aspects of life abroad. For instance, one interviewee remarks:

In general, they [the migrants] only show what is positive. They will show the beautiful landscapes and the beautiful clothes, but they will never show what they are working on, for example. (Modou, potential migrant, Dakar, 19 November 2019)

Second, respondents compare the social media posts to their own living situation, which makes the depicted reality even more desirable. For people who eventually migrate to Europe and are then experiencing a “reality check,” this often leads to great frustration. The realization that the living realities of Senegalese migrants in Europe tend to be quite different from what is depicted on social media can lead to a shock upon arrival. Idrissa, a migrant in Italy, explains his disappointment after he saw the real working conditions of Senegalese migrants with his own eyes:

The majority of them [i.e., Senegalese migrants in Europe] who came to Senegal [for holidays] or whose photos you saw, or who you found there, when you come here, when you see them working you say to yourself ‘no’. I'm not going to hide it from you, I was discouraged. […] You say to yourself that ‘no, it's impossible, it can’t be true, with those photos that he showed you, with what he made you believe, you know?’. (Idrissa, migrant, Rome, 19 September 2021)

Exploration of Mechanisms with Quantitative Data

Our qualitative analysis allows us to identify three potential mechanisms explaining the effect of the viewing of migrants’ social media content on migration aspirations. ¹² First, social media can facilitate one's cognitive migration (and thus, in turn, affect one's migration aspirations) by favorably shaping one's visions about the migration journey. Second, social media use correlates with an increased admiration of migrants abroad, making it more likely that one's own cognitive migration experience is positively shaped by such admiration, meaning that, e.g., interviewees may imagine themselves to be equally successful. Third, social media influences one's assessment of life in Europe through positively biased social media posts by migrants abroad. Interviewees are more likely to envision a pleasant and successful life in Europe during their cognitive migration, which can, in turn, fuel their migration aspirations.

To study whether these mechanisms also matter systematically, we use our survey data. We estimate the association between social media use (socialmedia) and various proxies for the potential mechanisms (transmission) for individual i, net of the influence of the usual vector of controls discussed above (X), using the following estimation equation:

t r a n s m i s s i o n j i = β 1 * s o c i a l m e d i a i + β ′ * X i ′ + ε i

(2a)

First, we expect social media content to facilitate the cognitive migration of individuals in Dakar and the Casamance by making the migration journey appear more viable. Individual perceptions of the migration journey are captured by answers to the questions “How likely is it that a migrant from Senegal makes it to Europe?” and “How likely is it that you will make it from Senegal to Europe?” For both questions, higher values mean that respondents are more likely to expect the journey to be successful. Second, the viewing of migrants’ social media content may increase one's admiration of migrants. Admiration of migrants is measured by answers to the questions “When you see successful migrants, does that encourage you or a suitable family member to migrate?” and “Have you or other members of your close family ever admired any other person who moved abroad?” Higher values coincide with more admiration. Third, social media content may make life in Europe appear more attractive and worthwhile, leading to a more pleasant cognitive migration experience. Individual perceptions about life in Europe are measured by answers to the questions “If a person from Senegal successfully migrates to Europe, how much money is this person able to send to the family in Senegal per month?” and “How often do you expect Senegalese migrants in Europe to be discriminated against?” For the former question, respondents will provide a sum of money as their estimate of potential remittances, while for the latter question, higher values coincide with higher levels of (expected) discrimination.

We expect the viewing of migrants’ social media content to coincide with higher admiration of migrants, more favorable imaginations about the migration journey, and greater expectations of life in Europe. Furthermore, we also study whether the six variables themselves share an association with individual migration aspirations. We consider the following model using the variables already described above.

m i g r a t i o n i = β 1 * t r a n s m i s s i o n j i + β ′ * X i ′ + ε i

(2b)

Our empirical results are reported in Table 4. They indicate that viewing more social media content by Senegalese in Europe correlates with a more favorable assessment of migration journey success and of life in Europe, as well as higher levels of admiration of migrants (Panel A). Assessing differences between Dakar and the Casamance, we find that effect sizes for social media use are similar across both sub-samples, with the exception of expected income, where social media use is only found to matter to the Casamance sub-sample (see Appendix E in the Supplementary Material for the full regression results for both sub-samples). Overall, the findings reported in Table 4 are consistent with our argument that social media use ultimately leads to stronger migration aspirations by shaping individual imaginations about migrants, migration, and life in Europe that facilitate one's cognitive migration.

OPEN IN VIEWER

Table 4.

Potential Mechanisms.

Panel A: association between social media use and potential transmission variables
	(1a)	(2a)	(3a)	(4a)	(5a)	(6a)
Dependent variable →	Migrant success 0–1	Admire migrants 0–1	Safe journey (others) 0–100	Safe journey (self) 0–100	Expected income (log)	Discrimination 0–3
Social media use	0.022***	0.035***	1.521***	1.179***	0.023*	−0.002
	(0.004)	(0.005)	(0.315)	(0.300)	(0.013)	(0.008)
Difference Dakar-Casamance χ2	0.86	0.42	0.01	0.71	4.51	2.13
(p-value)	(0.35)	(0.52)	(0.98)	(0.40)	(0.03)**	(0.14)
Further controls	Yes	Yes	Yes	Yes	Yes	Yes
Observations	3,471	1,823	4,037	4,037	3,817	4,037
R 2	0.051	0.043	0.038	0.020	0.028	0.015

Panel B: relationship between potential transmission variables and migration aspirations
	(1b)	(2b)	(3b)	(4b)	(5b)	(6b)
Transmission variable →	Migrant success	Admire migrants	Safe journey (others)	Safe journey (self)	Expected income	Discrimination
Transmission variable	36.684***	15.073***	0.238***	0.209***	0.976**	−3.648***
	(2.281)	(2.484)	(0.026)	(0.027)	(0.482)	(1.084)
Difference Dakar-Casamance χ2	3.03	0.40	64.14	43.11	0.03	7.74
(p-value)	(0.08)*	(0.53)	(0.00)***	(0.00)***	(0.85)	(0.00)***
Further controls	Yes	Yes	Yes	Yes	Yes	Yes
Observations	3,471	1,823	4,037	4,037	3,817	4,037
R 2	0.259	0.137	0.148	0.135	0.110	0.111

Notes: Dependent variable in Panel A is respondent's answer to the question “When you see successful migrants, does that encourage you or a suitable family member to migrate?” in Model (1a); “Have you or other members of your close family ever admired any other person who moved abroad” in Model (2a); “How likely is it that a migrant from Senegal makes it to Europe?” in Model (3a); “How likely is it that you will make it from Senegal to Europe?” in Model (4a); “If a person from Senegal successfully migrates to Europe, how much money is this person able to send to the family in Senegal per month?” (log of stated sum) in Model (5a); and “How often do you expect Senegalese migrants in Europe to be discriminated against?” in Model (6a). Social media use variable measures respondents’ social media exposure by their answer to the following survey question: “How often do you see photos or videos by Senegalese living in Europe on social media? With social media we mean Facebook, WhatsApp, Instagram, TikTok, Snapchat and similar apps or websites.” The answer range is from 0 (“never/practically never”) to 6 (“more than once a week”). Answer options in between are 1 (“less often than a year”), 2 (“once a year”), 3 (“once every three to six months”), 4 (“once a month”) and 5 (“once a week”). Dependent variable in Panel B is always the respondent's answer to the question “How much are you considering moving to another country to live?” (scale from 0 to 100). Further controls are for age, gender, education, economic expectations and relationship status. The reported χ²-statistic is associated with testing whether the effect size of social media use on the respective mediator (Panel A) or of the respective mediator on migration aspirations (Panel B) differs between Dakar and Casamance; an insignificant test statistic implies that effect sizes are indistinguishable between the urban and rural sub-sample in a statistical sense. To carry out this test, we run additional regressions of the models above for the Dakar or Casamance sub-sample, respectively. These regression results are reported in the Supplementary Material (Appendix E). Constant not reported. Heteroskedasticity-robust standard errors in parentheses. ***p < .01, **p < .05, *p < .1.

We also show that the various transmission variables correlate with migration aspirations in plausible ways (Panel B). For instance, we find that social media use increases the admiration of migrants; increased admiration, in turn, correlates with stronger migration aspirations. On closer inspection, there are some differences between Dakar and the Casamance (see Appendix E in the Supplementary Material for the full regression results for both sub-samples). Travel safety more strongly shapes migration aspirations in the Casamance, while concerns about discrimination in Europe matter more prominently for the migration aspirations of people from Dakar. Also concerning expected discrimination, while we find that this variable is negatively related to migration aspirations (especially for people from Dakar), there is no evidence that social media exposure correlates with individual discrimination expectations (Panel A). This latter result speaks to our qualitative findings that social media content produced by Senegalese in Europe is curated, i.e., biased towards positive images and imaginaries, and thus does not highlight experiences with discrimination.

As a robustness check, in the Supplementary Material (Appendix E), we also report results from a mediation analysis following Baron and Kenny (1986). Using the pooled data, according to this analysis, migrant success and migrant admiration, as well as the risk of oneself own death on the migration journey are more important mediators than expected income after migration and the migration risks for others. Using sub-sample data, we find that migrant success tends to matter more in Dakar, while concerns about the safety of the journey matter more to respondents from the Casamance. Consistent with our other findings, we again find that discrimination experiences in Europe do not mediate the effect of social media exposure on migration aspirations. Thus, the results from the mediation analysis largely conform to the other empirical results presented in this subsection.