Background & Summary

More than half of all tourism involves a coastal or marine destination, and blue tourism annually accounts for 4.6 trillion US dollar or 5.2% of the global gross domestic product1. Coastal destinations have been attractive for centuries because of their beneficial effects on mental and physical health2,3,4,5. Previous research found that spending time near the ocean reduces stress6,7,8, promotes physical activity9,10, and provides a setting to meet with family and friends11,12. It is now clear that residing in closer proximity to the coast and visiting the coast more often effectuates these benefits, similarly as what has been evidenced for inland blue and green spaces13,14,15,16. However, it is still unknown to which extent the characteristics of the coastal visit and visitor modify the health outcomes.

Studies from across the globe provided the initial evidence for the beneficial effects of the coast by revealing that residing in closer proximity to the coast is associated with a better self-reported general and mental health (e.g., in Belgium4, Canada17, China18, Ireland19,20, Japan21, Spain22, and the United Kingdom2,3,23,24,25). Several reviews and conceptual frameworks hypothesize that this pattern occurs because people who live nearer the coast tend to visit it more often16,26,27,28 (Fig. 1a). Indeed, cross-country analyses have confirmed that living nearer the coast is associated with a higher coastal visit frequency29,30 and that a higher coastal visit frequency is associated with a better self-reported general health5,15,30. It also seemed that these pathways are moderated by the demographic and socio-economic characteristics and health of the individual16,26,27,28. For example, Boyd et al.31 clearly illustrated for England that “infrequent users [of coastal environments] were more likely to be female, older, in poor health, of lower socioeconomic status, of ethnic minority status, live in relatively deprived areas with less neighbourhood greenspace and be further from the coast”. However, examples from outside England are now required to strengthen our understanding about how individual characteristics moderate the coastal visit frequency and experienced health effects.

Fig. 1
figure 1

Overview of what this dataset contributes to the current literature. Panel a shows a conceptual diagram of the mediating roles of visit frequency and visit characteristics and the moderating role of individual traits (ID) in explaining the benefits of residential coastal proximity for health. A distinction is made between what is known from the literature and what this data provides to complement existing knowledge. Panel b shows what this dataset contains.

Full size image

Next to the visit frequency, the visit characteristics may be an equally important mechanism by which residential proximity benefits general health. Coastal visitors perform many different leisure activities at the coast32, and each activity may result in distinct emotional, cognitive, and physical experiences that contribute to overall health33. Depending on the individual’s socio-demography and health, different activities may be performed and different health effects may be experienced32. Unfortunately, no study seems to have yet investigated who performs what kinds of recreational activities at the coast, and whether these different activities result in different experiences and health effects (Fig. 1a).

Knowledge Gap

There is a lack of high-quality scientific data that links individual characteristics such as residential proximity, demography, socio-economic status, and health to the coastal visit frequency and visit characteristics and the resulting emotional, cognitive, and physical experiences. Coastal nations’ tourism agencies usually do survey the activities, reasons, and experiences with the coastal amenities alongside the demographic features of their coastal visitors. However, in many cases, crucial variables such as residential coastal proximity, socio-economic status, and what people experienced emotionally and cognitively are neglected, because collecting the data often merely must serve the optimization of blue tourism. England seems to be an exception, because there the Monitor of Engagement with the Natural Environment (MENE) survey queries where people go, what they do, and what they experienced at the coast, alongside age, sex, deprivation indexes, and other characteristics of the individuals34. This data has proven to be very effective for science, because it has led to significant advancements in the current knowledge about coastal recreation and health relationships13,31,32,35,36.

Purpose of this dataset

The aim of this research was to develop a dataset that allows to perform confirmatory and exploratory investigations on the relationships between residential proximity to the coast and the resulting mental and physical experiences via the mediating effects of the visit frequency and characteristics, and the moderating role of the individuals’ socio-demography and health (Fig. 1b). Therefore, a survey was distributed among Flemish-speaking Belgian inhabitants and their visits to the Belgian coast (i.e. not international coastal tourism). We deliberately focused on a local scale to be able to reveal a diversity of relationships and patterns within the locally-specific cultural landscape. The data contains the visit frequency, visited municipal seaside resorts, performed activities, gained experiences, and applicable reasons for not visiting the coast more often alongside the geographic (i.e. postal codes), demographic, socio-economic, and health profiles of the respondents. We supplemented this data with the objective linear distances, shortest, and fastest driving distances from the centres of the home municipalities to the nearest and actually visited municipal seaside resorts, similarly as in previous research2,4. The dataset holds complete responses from 1939 respondents, of whom 1304 had visited the Belgian coast in the preceding year (67.25%), 627 had not visited the coast in the preceding year but did before (32.34%), and eight respondents had never visited the Belgian coast (0.41%). For the users’ convenience, each variable in the raw and processed data is given a detailed description in a codebook that is shared along with the data (folder ‘3. Processing’).

The dataset may be of interest for researchers aiming to disentangle relationships between the ocean and human health and for stakeholders of blue tourism37,38,39,40,41,42,43. More specifically, the data allows to unravel the social structuring of recreational activities to the coast, which can be analysed via multivariate modelling or ordination techniques. Alternatively, the data can also contribute to a number of ongoing investigations in the literature, and we propose thematic research questions that may be addressed with the data: about ‘coastal epidemiology and accessibility’, ‘health and psycho-physical experiences’, ‘social relations’, and ‘issues of time, season, and weather’ (Table 1). Four fields of application for the data are identified: to increase our understanding of the coastal recreation phenomenon; to help to address the needs, challenges, and opportunities in the blue tourism sector; to evaluate whether and how the coast can be used for new cost-effective health-care practices (e.g. coastal visits on prescription); and to help spatial planners to design the coast up to the needs of the residents and visitors. This publication provides univariate descriptions of the data, which can assist in shedding light on the variation present in the data and the quality of the data. By making this dataset publicly available and in accordance with FAIR principles, it also calls to researchers and tourism agencies to standardize coastal tourism questionnaires and make existing and newly acquired data openly available.

Table 1 A non-exhaustive list of thematic research questions that can be addressed with the proposed dataset. The themes may overlap with each other.
Full size table

Methods

Survey

The dataset contains the responses to an online survey about the performed recreational visits to the Belgian coast and the demographic, socio-economic, and health background of the respondents. The survey was distributed among a panel of 30.000 to 35.000 Flemish-speaking members from the five provinces in Flanders that had subscribed to participate in societally relevant research (Bpact, Leuven, Belgium). It was distributed from January 2nd to January 17th 2023 to meet the intended number of 1640 complete responses. Sampling happened via quota sampling based on data about the age (<34 y, 35–49 y, 50–64 y, 50–64 y and >65 y), sex, province, and educational attainment (categories: low, middle, high) that was previously gathered by the panel provider. Sampling happened during multiple waves while considering propensity scores per quota. Oversampling of quotas was allowed, and no exclusion criteria were set. In total, 2574 panellists responded to the survey, of whom 1939 provided a complete and reliable response (see section 4. Technical Validation). The respondents received points from the panel provider (quantity unknown for the researchers) for the time spent on the survey, and these points add up to an appropriate monetary compensation. The survey was anonymous and consent for voluntary participation was acquired via panel subscription. The research was conducted according to the ethical rules presented in the General Ethical Protocol of the Faculty of Psychology and Educational Sciences of Ghent University. The survey was administered in Dutch via the online Bpact user interface and Qualtrics software44. The survey itself can be found within the data45 in the folder ‘/1. Survey’ and the acquired responses in folder ‘/2. Raw data/a. Survey responses/’.

Asking about respondents’ socio-demographic background and health could lead to social desirability bias. To be able to assess potential measurement error, we asked at the end of the survey how comfortable the respondents were with answering each section of the survey (i.e. about coastal visits, demography, employment situation and income, and health) using a five-item multiple choice with answers ‘very discomfortable’, ‘discomfortable’, ‘neutral’, ‘comfortable’, and ‘very comfortable’46.

Coastal visits

The questions about the coastal visits in the survey were designed to optimally capture the diversity of visit frequencies and characteristics. Furthermore, the questions meant to capture the respondents’ general perceptions and trends about many of their past coastal visits across seasons and years, rather than detailed information about only a couple of their visits (e.g., of visits in the last four weeks, as in previous research15,29). These general perceptions and trends were deemed to be more indicative for summarizing a respondent’s coastal visit behaviour and for distinguishing visit profiles across socio-demographic groups. Coastal visits were operationalized as days at which the person was at the Belgian coast in a recreational context and saw the sea. Depending on whether the respondents’ visited the Belgian coast in the previous year or before that, the reference period differed and additional questions about the frequency and locations of the performed visits were asked (Table 2). All respondents were asked to report how frequent 32 activities were (or would be) performed, how frequent the person was (or would be) accompanied by 7 types of social company, how frequent 27 experiences were (or would be) felt, and how frequent 18 reasons for not visiting the coast more often applied (or would apply; Table 2). Response categories for these questions were ‘never’, ‘seldom’, ‘sometimes’, ‘often’, and ‘always’. The items and response categories of the activities, types of social company, experiences, and reasons for not visiting the coast more often were chosen based on the local culture, the potential outcomes and mechanisms described in the nature and health literature16,28,47,48, previous studies about the experiences along the Belgian coast49,50, and the following previous surveys: the ‘Monitor of Engagement with the Natural Environment’ survey being administered nationally in England (MENE)32,34, the ‘Cultural participation in Flanders’ surveys being administered yearly in Flanders (from 1996 as the ‘SCV-survey’51,52,53, and since 2019 as part of Flanders’ ‘SV-survey’), and the surveys administered to day visitors and stayers in coastal accommodations by the local tourism agency aimed at informing policy54,55. Thus, we did not blindly copy item sets that were available in the literature, but rather designed our own based on our perspectives on the current knowledge. The respondents who visited the coast in the previous year also had to report for each of the four seasons within that year how many days the person was at the coast during a coastal day visit or during a multi-day visit with overnight stay in an accommodation at the coast. They also had to report which of the 14 municipal seaside resorts were visited. The respondents who had not visited the coast in the previous year but did before had to additionally report the frequency of coastal visits in that period, and which of the 14 municipal seaside resorts were then visited. Figure 2 displays the variation in the responses with regard to the coastal visits.

Table 2 Overview of the questions about the visits to the Belgian coast that were asked to the respondents who visited the Belgian coast in the previous year, to those who did not visit the Belgian coast in the previous year but visited the Belgian coast before that, and to those who never visited the Belgian coast.
Full size table
Fig. 2
figure 2

Descriptive graphs of the queried coastal visit frequency and coastal visit characteristics. Panels a to d show the acquired information about the visit frequency, panels e to i about the visit characteristics.

Full size image

Note that the Belgian coast consists of 10 administrative municipalities, but the names that Belgian citizens commonly give to the 14 municipal seaside resorts often differ from the administrative boundaries (Table 3, Fig. 3).

Table 3 The names of the municipal seaside resorts as queried in the survey and the administrative sub-municipalities to which they respectively refer to.
Full size table
Fig. 3
figure 3

Map of the Belgian coast with municipal seaside resorts queried in the survey. Municipality borders are delineated by a solid line, the borders of sub-municipalities to which the seaside resorts belong by a dashed line. Open black circles represent the centroids of the polygons of the municipal seaside resorts, solid black circles the modelled destinations.

Full size image

Demographic background

The section in the survey about the demographic background aimed to capture the life stage and living situation in physical and social space. The queried traits of the respondents were the following: year of birth, gender, the postal code of the primary residence, four optional postal codes of secondary residence locations, whether the respondent had a secondary residence at the coast that is sometimes visited for leisure, the perceived residential distance to the nearest coastline, the number and types of co-inhabitants, the number of people on which the respondent can count on when faced with serious problems (this self-designed proxy for social support is a simplified version of a previously-published single-item questionnaire for social support56 and hints to the same constructs as a multi-item social support questionnaire57), whether and where the person grew up in Belgium, and how often the coast was visited during childhood. The user can link these data to the frequency and characteristics of the coastal visits to evaluate the influence of geography, the social context, and mechanisms of nostalgia – an emotion that has proven to be crucial when investigating coastal visits and experiences50,58. Figure 4 panels a, b, c, f, j, and k visualize the variation in these demographic parameters. Some respondents’ postal codes (1000 and 1090) were in the Brussels-Capital Region (N = 2) and were not meant to be sampled via the panel, but we kept these respondents in the data for completeness.

Fig. 4
figure 4

Descriptive graphs of the demography, socio-economic status, and health of the respondents. Panels a, b, c, f, j, and k show the demographic and social context of the respondents, panels d, e, g, h, i, and l show the socio-economic context, and panels m, n, o, and p their health and physical activity.

Full size image

Socio-economic status

The survey included different proxies for the respondents’ socio-economic status. Firstly, the educational attainment was queried using the descriptions of the nine main categories in the International Standard Classification of Education (ISCED 0 to 8). There was also an ‘other’ category where respondents could specify their educational degree in case of uncertainty, but this resulted in some unclear responses that were identified as ‘NA’. Secondly, the employment situation distinguished the active (i.e. employee, self-employed with and without employees, student full-time education) from non-active (i.e. unemployed, retired, housewife or homemaker, out due to sickness or other circumstances) population using a multiple-choice question with one possible answer. There was also an ‘other’ category. The employees and self-employed respondents were also asked about their employment time (i.e. working full-time or part-time) and their occupation using the first-order and second-order classifications of the International Standard Classification of Occupations (ISCO). The survey also queried how often the respondent’s occupation involved being at a coastal or marine environment to evaluate the potential of constrained restoration59,60,61,62. Lastly, all respondents were asked about their net household income using increments of thousands (i.e., <1000 euros/month, 1001–2000 euros/month, …, >6001 euros/month). Figure 3 panels d, e, g, h, I, and i visualise the variation in the socio-economic responses.

Health

Information about the health of the respondents was gathered for three reasons. First, it could help evaluate whether coastal visit behaviours are moderated by a person’s mental and physical health, for example in cases of limited mobility or depressive symptoms with a tendency towards social isolation. Second, it could help identify whether particular coastal visit behaviours, such as visiting the coast more often or performing particular coastal activities, are associated with a better or worse health as an outcome. Thirdly, it could provide further support for the relationship between living nearer the coast and self-reported general health or other proxy for health4. The questions related to the health of the respondents included the self-reported general health (first item from the short-form health survey, SF1), aspects of mental health (mental health part of the short-form health survey, SF36MH), including items referring to arousal/vitality (liveliness, vitality, exhaustion, tiredness) and to the valence/emotionality (nervosity, depression – feeling troubled, depression – feeling down, happiness), and having been diagnosed with a mental or physical chronic condition and whether this condition is still present. Also questioned were the time spent doing light, moderate, and intense physical activity in the past month (international physical activity questionnaire short form; IPAQ-SF)63. Figure 3 panels m-p visualize the variation in the health of the respondents.

Nature connectedness

The survey was closed with a one-item question about the respondents’ nature connectedness, stating “Do you rather agree or disagree with the following sentence? Like a tree can be part of a forest, I feel embedded within the broader natural world”. Answers categories were ‘totally agree’, ‘agree’, ‘neutral’, ‘disagree’, ‘totally disagree’. This question is one item from the connectedness to nature scale that has shown to be particularly indicative of nature connectedness based on item response theory64.

Processing steps

The processing steps with regard to the survey data can be found in Tables 4 and 5.

Table 4 Listed questions for the coastal visit characteristics in the survey and the type of answers and processing steps. Besides the mentioned processing steps, every variable was also translated from Dutch to English.
Full size table
Table 5 Listed questions for the individual characteristics in the survey and the type of answers and processing steps. Besides the mentioned processing steps, every variable was also translated from Dutch to English.
Full size table

Additional calculations

Weights

The quota sampling inevitably caused an imperfect representativeness of the sample for the population. To clarify these sampling errors and correct for them to a possible degree, post-stratification weights were calculated for different strata based on the combinations of age (18–29 y, 30–39 y, 40–49 y, 50–59 y, 60–64 y, >= 65 y, total), sex (male, female, total), educational attainment (low, middle, high, total), and province (West Flanders, East Flanders, Antwerp, Limburg, Flemish Brabant, total). For each stratum, the weight was calculated as the frequency of individuals that belong to the stratum in the population divided by the frequency of individuals that belong to the stratum in the sample. Population statistics were retrieved from Statbel65. The number of strata were reduced to retain the specificity of the strata while limiting the number of excessively high or low weights due to exceptionally under- or oversampled strata, respectively. More specifically, consecutive age categories that were either both under- or both over-represented by the sample were pooled together (Fig. 4 panel a), so that the deviations from representativeness did not cancel out during the pooling. Since the population data contained sex data, and not gender data as in the survey, we had to consider the gender categories ‘Male’ and ‘Female’ to be matching the sexes ‘Male’ and ‘Female’ from the population. This gender-sex linkage and the different meanings of these concepts may have resulted in misrepresentative weights. Educational attainment was pooled into ‘Low’, ‘Middle’, and ‘High’, because the quota sampling also adopted these categories and this considerably decreased the number of excessively high weights. The Province categories were not adjusted to retain geographical specificity for each weight. If information for these four parameters were incomplete for an individual (e.g. NA for Educational attainment), those parameters were disregarded and the weight was calculated based on the parameters for which information was available. There were 504 strata with different combinations of age, gender-sex, educational attainment, and province categories. The weights were trimmed at 0.2 and 5. The population statistics retrieved from Statbel65 can be found within the data45 in the folder ‘/2. Raw data/b. Population socio-demographics/’. The weights table with pooled ISCED and age categories as described above is stored within the data45 on location ‘/3. Processing/R workspace/Weights.ISCEDLMH.Age.csv’. All weights were appended to the final processed survey data.

Residential proximity to the coast

Residential proximity to the coast was operationalized in different ways depending on the type of residence (primary and secondary; in 2022 and as a child), the destination (nearest coast vs. visited seaside resorts), and the route between the two (linear, shortest and fastest driving route; Fig. 5). Firstly, the distances were calculated from all residential primary and secondary postal codes in 2022 and as a child to the nearest coast. Secondly, the distances were calculated from the residential primary and secondary postal codes in 2022 to the actually visited seaside resorts in that year. As such, the user of this dataset can choose the proxy that is best suited for quantifying the residential proximity to the coast. For each distance, the residential municipality centroids were used as starting point. The distances corresponding with the linear and shortest and fastest driving routes were derived in kilometres, and for the shortest and fastest driving routes also in the number of seconds travel time. The destination points of the visited seaside resorts were the points along the shorelines that were closest to the centroids of the municipal seaside resorts (Fig. 3). The routes and accompanying distances were calculated with geographical information system (GIS)-methods: QuantumGIS 3.2.2 was used to generate a map with the OpenStreetMap road network (OpenStreetMap contributors, 2018) and Eurostat coastline data (Nomenclature of Territorial Units for Statistics (NUTS), 2013) and the ArcGIS Pro 2.2.0 Network Analyst extension was used to generate the routes and calculate the distances. Next to these objective measures of residential proximity, the perceived residential proximity to the nearest coast (in kilometres) was also queried in the survey. The raw distances calculated via GIS can be found within the data45 in the folder ‘/2. Raw data/c. Residential distances to the coast/’. The distances to the nearest coast were appended to the survey data with R66, which is stored at ‘/4. Processed data/’ in the files ‘Hooyberg_Survey_Processed.csv’ and ‘Hooyberg_Survey_Processed.txt’. The distances to the destinations were stored in a separate dataset due to its long format in the folder ‘/4. Processed data/’ in the files ‘Hooyberg_Distances_Processed.csv’ and ‘Hooyberg_Distances_Processed.txt’. The dataset also holds a merged dataset with all survey data and distances under the folder ‘/4. Processed data’/ in files ‘Hooyberg_Survey_Distances_Processed.csv’ and ‘Hooyberg_Survey_Distances_Processed.txt’.

Fig. 5
figure 5

Visualization of the proxies for residential proximity to the coast. Panel a displays the sampled municipalities and the fastest driving routes to the coast. Panel b shows an example of three types of residences of a respondent. Panel c illustrates that the distance can be calculated to the nearest coast and to the actually visited seaside resorts from the primary postal code in 2022 from a respondent. Panel d displays an example of the different types of routes and associated distances and travel times to the nearest coast. Panel e shows the summary of the respondents’ perceived residential distance to the nearest coast.

Full size image

Data Records

To provide the reader with maximal transparency about the questioning of the items in the survey and the processing steps implemented for the final datasets, the zipped data contains four folders with the survey itself, the raw data, the processing process, the processed data, and the figures. All data can be found within the data45. The zipped file is 161 megabytes in size and the unzipped file 372 megabytes.

Folder ‘1. Survey’

The first folder provides the survey in different formats and languages. The survey was administered in Dutch using the Qualtrics survey administration software, and English translations were added in Qualtrics later for publication.

  • The file ‘Hooyberg_SurveyCoastalVisits_NL.docx’ is the exported human-readable format of the survey with all the rules and flows of the survey in the original language Dutch.

  • The file ‘Hooyberg_SurveyCoastalVisits_ENG.docx’ is the same but with the English translation (translation was done after administration).

  • The file ‘Hooyberg_SurveyCoastalVisits_NL_print.pdf’ clearly visualizes the layout and how the survey was shown to the respondents.

  • The file ‘Hooyberg_SurveyCoastalVisits_Qualtrics.qsf’ is the Qualtrics project that can be loaded into the software.

  • The file ‘Hooyberg_SurveyCoastalVisits_translations_NL-EN.csv’ contains the Dutch to English translations of the survey, which was downloaded from Qualtrics and which can be uploaded again if translations would have been lost in the Qualtrics project.

Folder ‘2. Raw data’

The raw data folder contains the raw survey responses (folder ‘a. Survey responses’), the population statistics from which the weights were calculated (folder ‘b. Population socio-demographics’), and the proxies for residential proximity to the coast derived by GIS (folder ‘c. Residential distances to the coast’).

Folder ‘a. Survey responses’

The folder ‘a. Survey responses’ contains the complete and incomplete responses that were downloaded from Qualtrics at the end of the survey administration on January 18th 2023. Each folder has the responses in different formats: comma separated values format (.csv), a format to be loaded in to IBM SPSS Statistics (.sav), tab-separated values format (.tsv), MS Excel standard format (.xlsx), and the MS Excel-compatible extensible mark-up language format (.xml).

Folder ‘b. Population socio-demographics’

The folder ‘b. Population socio-demographics’ contains the original population statistics per stratum as originally received by Statbel (file ‘Hooyberg_Pop_2023-08-01-Statbel.xlsx’) and the re-formatted data in wide (file ‘Hooyberg_Pop_2023-08-10_wide.xlsx’) and long (file ‘Hooyberg_Pop_2023-08-10_wide.xlsx’) format to be loaded into R.

Folder ‘c. Residential distances to the coast’

The different proxies for residential proximity to the coast are stored in two files. The first file (‘Hooyberg_DistancesToNearestCoast.csv’) contains a list of all the residential postal codes reported by the respondents and the corresponding distances to the nearest coast. The second file (‘Hooyberg_DistancesToDestinations.csv’) contains for each respondent the distances from the different types of residences to the visited seaside resorts. Both files were later merged with the survey data (see also section Methods – Additional calculations - ‘Residential proximity to the coast’).

Folder ‘3. Processing’

The processing folder contains the codebook and the folder ‘R workspace’. The codebook (‘Hooyberg_SurveyCoastalVisits_Codebook.csv’) describes all of the original and newly added variables in the data with their coded names and formats. The folder ‘R workspace’ contains the latest version of the R script (named ‘Hooyberg_SurveyCoastalVisits_2023-08-31.R’), the weights calculated from the population and sample statistics per stratum (‘Weights.ISCEDLMH.Age.csv’, see section Methods – Additional calculations - Weights), and the postal codes for the municipal seaside resorts (‘Seaside_resorts_ZIP.csv’). The R workspace folder also contains the folder ‘Adjustments’. This folder contains help files for correcting respondents’ erroneous answers to the ‘other’ answer categories for household company (‘Household_Company_other_adjustments.csv’), education level (‘Education_other_adjustments.csv’), and occupational employment (‘Employment_other_adjustments.csv’). The folder ‘Adjustments’ also contains the categorisations and translations of the chronic mental and physical illnesses reported by the respondents (‘Chronics_mental_and_physical_illness_translation.csv’). Lastly, the ‘Adjustments’ folder contains two files that were used to identify the invalid responses by speeding and/or straightlining. A first file contains all the complete responses (N = 1949) and was used to visually scroll through the data to search for patterns of speeding and/or straight-lining (‘Survey_speeders_straightliners_highlighted.xlsx’). A second file contains the ID’s of these invalid responses (N = 10) that were loaded into R for exclusion from the data (‘Survey_speeders_straightliners_IDs.csv’).

Folder ‘4. Processed data’

The folder ‘4. Processed data’ harbours the final survey data in which the distances to the nearest coast are embedded (files ‘Hooyberg_Survey_Processed.csv’ and ‘Hooyberg_Survey_Processed.txt’). It also contains the final distances data to the visited seaside resorts (files ‘Hooyberg_Distances_Processed.csv’ and ‘Hooyberg_Distances_Processed.csv’). Both of these datasets were also merged together (file ‘Hooyberg_Survey_Distances_Processed.csv’ and ‘Hooyberg_Survey_Distances_Processed.txt’).

Folder ‘5. Figures’

The folder ‘5. Figures’ contains the original figures included in this descriptor and their individual panels. The panels were imported in Adobe Illustrator (file ‘Hooyberg_Survey_Descriptive_Graphs_2022-08-31.ai’) and their format was clarified and made consistent for resulting in the final files for Fig. 2, Fig. 4, and Fig. 5. The folder ‘Raw figures from GIS’ holds Fig. 3 and all the original panels of Fig. 5 that were generated via GIS. The folder ‘Raw figures from GIS’ also holds a figure overviewing all the primary and secondary residential postal codes in 2022 and those as a child. The folder ‘Raw figures from R’ contains the original panels of Fig. 2 and Fig. 4 that were generated and exported in R based on the survey data.

Technical Validation

Representativeness for the Flanders’ population

The use of an access panel greatly increased the representativeness of the Flanders population compared to convenience or similar other survey sampling methods. It also ensured that all invited panellists were blinded for the survey topic to reduce selection-bias. However, the sample was not perfectly representative for the population based on age, gender/sex, educational attainment, and province of residence. In general, the majority of the strata were oversampled and the majority of the sampled individuals (77.6%) belonged to an oversampled stratum with a weight less than 1. This was to be expected from the quota sampling procedure with oversampling allowance performed by the chosen panel provider. As a result, the data contains an overrepresentation of individuals that are 60-to-64-year-old, 40-to-49-year-old, belong to the middle socio-economic class, and reside in another province than Flemish Brabant (Fig. 4 panels a and e). Underrepresented are the individuals that are 50-to-59-year-old, of lower socio-economic class, and who reside in Flemish Brabant (Fig. 4 panels a and e). Also interesting was that 1 stratum had a weight of lower or equal than 0.2 that was assigned to 42 respondents (2.17%). Twenty-five strata had a weight higher or equal than 5 that was assigned to 5 respondents (0.26%). The data provided by the panel provider was insufficient to calculate design or non-response weights. It is difficult to compare the data about the coastal visits with other sources (e.g. by tourism agencies) because of the different sampling designs, different reference periods, and units of measurement.

Response quality

All original questions (in Dutch) with their best possible English translations can be found within the data (folder ‘1. Survey’), and any issues with regard to the quality of the responses can be attributed to the manner of asking the questions46,67.

Only complete and valid responses were retained (N = 1939). Responses were regarded as complete when the last question of the survey was answered (N = 1949). Respondents that subsequently not proceeded to the end page or to the BPact panel user interface (N = 86) were retained, but note that these responses were not considered during the quota sampling and may have resulted in disproportional oversampling of the quotas (but see section Methods – Additional calculations – Weights and section Technical Validation – Representativeness for the Flanders’ population). Responses with relatively quick answering patterns (‘speeders’) and with repeated similar – often contradictory – answers (e.g. ‘always’ on all of the performed activities; ‘straight-liners’) were identified as invalid and were disregarded (N = 10). These invalid records were identified by visually searching through the data for records with the same responses throughout the survey. This process can be retraced in the file ‘Survey_speeders_straightliners_highlighted.xlsx’ in the folder ‘3. Processing/Adjustments’. We did not specify a cut-off on response times because case-by-case evaluation of the data by a researcher was more informative about the response quality. Section ‘Methods – Processing steps’ further provides what corrections were done to enhance the quality of the responses.

The proxies for residential proximity to the coast are based on residential postal codes, and not on accurate coordinates or addresses. As such, the user should keep in mind that the linear, shortest, and fastest travel routes reported in the data probably differ to a certain degree from the real routes.

Usage Notes

Familiarization with the survey design, questions, response options, and processing steps performed is encouraged before interpretation, exploration and analysis of the data. After familiarization, the final processed dataset can be explored and analysed with the desired statistical software at any difficulty level to answer any of the research questions proposed in the introduction or other ones. Ideally, the survey weights are to be considered during the analyses, and the scope in time (i.e. 2022) and space (i.e. Flemish inhabitants and visits to the Belgian coast) should be respected during interpretation.