THE UNEVEN IMPACT OF THE COVID-19 PANDEMIC ON DOMESTIC TOURIST FLOWS: WHAT DOES MOBILE PHONE DATA TELL US? Author Details: Ana Condeço-Melhorado*1; Juan Carlos García-Palomares2; Javier Gutiérrez3 * Corresponding author 1 Ana Condeço-Melhorado. Departamento de Geografía, Universidad Complutense. ORCID 0000-0002- 7260-3260 2 Juan Carlos García-Palomares. Departamento de Geografía, Universidad Complutense. ORCID 0000- 0002-8759-6809 3 Javier Gutiérrez. Departamento de Geografía, Universidad Complutense. ORCID 0000-0003-2611-8587 Structured Abstract: Purpose: The COVID-19 pandemic has significantly impacted global tourism, with international travel bearing the burden of restrictions. Domestic tourism has also faced substantial challenges. This paper analyses the impact of the COVID-19 pandemic on domestic tourism in Spain, focusing on travel from Madrid (the country's capital) to other tourist destinations. Design/methodology/approach: Mobile phone data has been used to study the evolution of tourist trips over the summers of 2019, 2020 and 2021. Regression models are used to explain the number of visitors at destinations. Findings: The pandemic not only caused a drastic drop in tourist flows, but also disrupted the overall pattern of the domestic flow system. Winning destinations were typically areas in proximity to Madrid and less densely populated destinations, while urban destinations were major losers. The preferences of domestic tourists varied notably by income group, but the decrease in trip volumes showed only marginal differences. Originality/value: The paper demonstrates the potential of mobile phone data analysis to study the uneven impact of external shocks, such as the COVID-19 pandemic, on tourist destinations. This approach considers spatial resilience heterogeneity within regions or provinces. By incorporating income information, the analysis introduces a social dimension to highly detailed spatial data, surpassing traditional studies conducted at the regional or national levels. KEY WORDS. Domestic tourist mobility, COVID-19 pandemic, mobile phone data, Spain INTRODUCTION The COVID-19 pandemic reshaped tourist trips and destination preferences, with a significant decline in travel compared to daily and weekend mobility (Edsberg Møllgaard et al., 2022). Sealed borders, event bans, closed attractions, and mobility restrictions globally affected domestic and international travel (Arbulú et al., 2021). The unprecedented impact of the health crisis on tourism is evident in both the volume of the affected population and the global scale (Yang et al., 2022). COVID-19 severely impacted Spain's economy, with tourism constituting 12% of GDP. Restrictions, starting on March 14, 2020, led to a 77% drop in foreign tourists in 2020 (83.6 million to 19 million). Summer 2020 saw eased domestic travel between regions, fostering local tourism amid an overall trip reduction. Google and Facebook data indicated a significant decline in trips that summer (Pérez-Arnal et al. 2021). By summer 2021, international tourism had partly recovered, but domestic tourism rebounded even more. Most COVID-19 impact studies on tourist destinations use provincial or regional data (Boto- García and Mayor, 2022; Duro et al., 2022; Falk et al., 2022) or focus on specific destinations (Ren et al., 2022; Yu et al., 2023). Notably, none address the pandemic's spatially disaggregated impact on tourist mobility within a country. This is crucial as different destinations within the same region or province may respond differently to shocks like COVID-19. There's a lack of empirical evidence on these mobility variations, hindering policymaking for crisis management (Yu et al., 2023). Our paper fills that research gap by using mobile phone data to analyse the uneven impact of the COVID-19 pandemic on tourist flows within regions or provinces. We focus on Madrid residents' trips to other Spanish destinations in 2019, 2020, and 2021, examining spatial heterogeneity. Additionally, we enrich mobile positioning data with census tract-level income data, revealing income-related variations in tourist behaviour and pandemic-induced changes in mobility. This social dimension, often studied via surveys (Menor-Campos et al., 2020) and with less spatial detail, is analysed here using the location coefficient approach. Analysing flow volume changes, destination effects and social impacts at a finer level of detail is essential for understanding external shock impacts on domestic tourism and devising equitable recovery strategies. The structure of this paper is as follows: after this brief introduction, section 2 reviews the related literature, section 3 presents the characteristics of the data and the methodology, and section 4 shows the results. Finally, in the discussion and conclusion section, some theoretical and practical implications are presented. RELATED LITERATURE Tourist mobility and the public health crisis: the outbreak of COVID-19 Global tourist flows face disruptions by external shocks, such as natural disasters, political instability (Causevic & Lynch, 2013), terrorist attacks (Corbet et al., 2019), or epidemics (Prideaux, 2005). The impact is complex depending on their nature, magnitude, duration, and affected areas (Backer & Ritchie, 2017). Furthermore, crises may repeat over time with different frequencies, posing increasing public health travel risks due to the rise of tourists numbers, longer distances travelled (Cui et al., 2016). Perceived risk, especially health-related, can affect travelers' destination choices and behaviour (Golets et al, 2023). The COVID-19 pandemic heightened tourists' perceptions of health risks, negatively affecting their travel intentions (Neuburger & Egger, 2021) and destination choices (Timur et al., 2023 and Liu and Zheng, 2023), impacting territories differently depending on demand, supply or mitigation/adaptation policies (Duro et al., 2021). Government mobility restrictions and public risk perception led to a significant decline in tourist flows (Cerón et al., 2021; Bae et Chang, 2021), revealing variations in vulnerability and resilience among destinations. Traveling abroad was considered more risky than domestic trips, rendering remote destinations, like Caribbean islands, more vulnerable and less resilient than national destinations. Likewise, nature tourism experienced unprecedented popularity and growth, while crowded urban destinations were perceived as high contagion risks (Viana-Lora et al., 2023). Despite this, certain mass tourism destinations, like beaches, attracted substantial tourist numbers (Zielinski and Botero, 2020). The pandemic crisis was considered an excellent opportunity to rethink the tourism model and move towards more sustainable tourism (Gössling et al., 2020). However, post-pandemic, tourism displayed resilience, with pre-crisis tourist numbers already rebounding. Emerging data sources, tourist mobility research and COVID-19 Technology is instigating profound changes in the tourism sector, giving rise to Smart Tourism - an emerging field in tourism research (Johnson & Samakovlis, 2019; Mehraliyev et al., 2019). This term describes the increasing reliance of tourism destinations, industries and tourists on emerging forms of ICT that transform data into value propositions (Gretzel et al., 2015). Smart Tourism Technologies include online tourism applications and information sources such as online travel agents, blogs, websites, social media, and smartphones, providing correct information, better decision support, increased mobility and quality tourism experiences to consumers and service providers (Azis et al., 2020). Smart Tourism relies on capturing, storing and analysing massive data, with fixed and mobile sensors, smartphones, GPS, bank cards, travel cards or social media monitoring tourist’s spatial behaviours throughout the day. During the phases of tourist trip - travel to site, onsite experience and return - tourists leave their digital footprint (García-Palomares et al., 2015), encompassing pre-trip searches, online reservations, onsite activities, social media sharing, and post-trip reviews. This amount of data, related to the 4 As of tourism infrastructure - Accommodation, Access, Amenities and Attractions – is crucial for destination development (Haneef, 2017). Smart tourist destination managers leverage this data intensively to improve the destination’s positioning amid increasing competition (Baggio et al., 2020). Big Data is excellent raw material for identifying temporal and spatial patterns in tourist flows (Chantre-Astaiza et al, 2019). Following the digital footprint left by tourists, we can analyse tourist mobility based on their spatio-temporal trajectories. Examples include mobile phones records (Kuusik et al., 2011; Raun et al., 2016), social networks (Barros et al., 2020) or bank card payments (Sobolevsky, 2014). Other technologies, such as Bluetooth tracking (Versichele et al., 2012) and GPS devices (Shoval and Ahas, 2016), have also been employed to analyse the spatio- temporal behaviour of tourists. Mobile phone data have clear advantages in the analysis of spatio-temporal patterns of tourists. Call Detail Records (CDR) (call and message logs), Data Detail Records (DDR) (internet access logs) and positioning data from the regular updating of devices by the network of mobile phone towers, enable the identification of user location over time (typically updated hourly). This allows inference of spatio-temporal trajectories. They provide large sample sizes (millions of registers), and extensive coverage, including less visited places (Ahas et al., 2007). In the field of tourist mobility, mobile phone data have been used to analyse spatial patterns and seasonal variations of foreign tourists by nationality (Ahas et al., 2007), destination loyalty through the identification of repeat visitors (Kuusik, et al. 2011; Tiru, et al., 2010), travel pattern comparisons between occasional and regular visitors (Nilbe, et al. 2014), and spatio-temporal patterns of tourist activities (Raun et al., 2016). Recent studies have delved into tourist flows within destinations. Thus, Wang et al., (2021) analysed the intra-urban tourist flow structure, constructing a trajectory network of tourists. Qian et al. (2021) studied tourist behaviour focusing on activities and interactions within facilities, accommodation and transportation hubs. Xu et al. (2021) analysed international travellers' trajectories to derive indicators such as the duration of stay and the spatial activity scope. Finally, Chu and Chou (2021) constructed a trajectory network of tourists in a tourist region, identifying the functions of each destination. Studies on the impact of the COVID-19 pandemic on tourist mobility used different approaches and data sources. Arbulú et al. (2021), Moreno-Luna et al. (2021), Boto-García & Mayor (2022), Duro et al. (2022), and Falk et al. (2022) analysed the non-uniform impact of the pandemic on tourist destinations in Spain and France using official data sources on travellers or overnight stays at the regional/provincial level, without considering the wide variety of destinations within each region/province. Pérez-Arnal et al. (2021) employed Facebook, Google and Apple data for research on tourist mobility in Spain during the pandemic at the regional level. While these studies did not show unanimous results, they observed that rural tourism suffered less from the COVID-19 shock, while islands were more affected. Other researchers explored different perspectives on tourist mobility during COVID-19 pandemic. Ren et al. (2022) used Navigation Data for studying domestic travellers in a Korean city by trip purpose. Kupfer et al. (2021) analysed foot traffic data collected from mobile devices for National Park visitors. Gao et al. (2021) studied urban tourist behaviour using social media data, and Yang et al. (2022) used Google Destination Insights for analysing international tourist flows. Woo-Hyuk et al. (2023) looked at COVID-19’s impact on travel distances to hotels and intraregional variances at different pandemic stages. The study conducted by Yu et al. (2023) is one of the few studies looking at the impact on domestic tourist flows using mobile phone data. Their study examines domestic tourist flows from Chinese cities to Beijing, finding a greater impact on large cities. In contrast, our study reverses this approach, analysing domestic tourist flows from Madrid to all Spanish destinations, providing a detailed spatial analysis over 2019- 2021, a unique approach in the existing literature. DATA AND METHODS Mobile phone data and software Mobile phone data for tourist trips from Madrid to the rest of Spain was obtained through a collaboration with Kido Dynamics, a data intelligence company. Kido Dynamics utilized anonymized records from a major Spanish telecommunications operator, representing over 20% market share and diverse social groups. The data, generated when mobile phones interact with the network (CDR and DDR), indicate the location of connected towers rather than users' exact locations. Tourist trips from Madrid to various domestic destinations were analyzed for August in 2019 (pre-pandemic reference), 2020, and 2021. August is the peak summer holiday month in Spain. Tourist trips are trips that last for at least two overnight stays at the destination. Maps and most of the data analysis were produced using a Geographic Information System (ArcGIS Pro 3.1). Bar charts, bivariate correlations and regression analysis were performed on MS Excel and SPSS. Spatial aggregation Once processed, the data were upscaled to represent the entire population of Madrid, and aggregated in origin and destination trip matrices. Spatial referencing of the mobile phone towers was aligned with administrative units. Origins comprised census tracts in Madrid, grouped for anonymity into 179 contiguous tract clusters. Destinations, were municipality- based or grouped for sparsely populated areas, resulting in 650 destinations. Both origins and destinations maintained high spatial resolution. The origin and destination trip matrices recorded 116,474 in 2019, 91,674 in 2020 and 102,601 in 2021, involving at least one trip. Enrichment of tourist mobility data by income level Trip matrices were enriched by incorporating average income per person data from the Spanish National Institute of Statistics at the census tract level. This information was linked to the origins and destinations of the Origin-Destination (OD) travel matrix. Given that these origins/destinations are aggregations of census tracts, the income data was assigned by first calculating the average income for each origin/destination. Populations-weighted averages of each census tract income within each origin/destination were computed to ensure accurate representation. The origin zones were classified into four income levels using an ad-hoc method, using quartiles for a balanced distribution. The classification is based on net per capita income, with cut-off values set at: low (below €12,308 per year), lower-middle (below €16,143), upper-middle (below €21,247) and high (above €21,247). According to this classification, 30% of Madrid’s population resides in low-income areas, while nearly 60% inhabit areas with incomes below the median (Table 1). Table 1. Areas of origin according to income level and population. Income per capita (euros) Population Quartile Number of areas Mean Std Dev Total % Q1 (< 12,308) 45 10,205 1,771.5 1,049,198 34.3 Q2 (12,308-16,143) 44 13,946 1,058.9 787,823 25.8 Q3 (16,143-21,247) 45 18,691 1,482.9 731,114 23.9 Q4 (> 21,247) 45 25,798 307.0 491,013 16.1 Total 179 17,178 6,177.3 3,059,148 100.0 Source: Authors own creation Descriptive analysis of the tourist trip matrices Based on the travel matrices, an analysis of tourists distribution by destination was conducted, mapping total trips in August 2020 and 2021, along with the deviation from the baseline (2019). The analysis delves into the evolution of main tourist destinations, aggregating data by destination type. This has allowed us to identify pandemic impacts on travel patterns between coastal and inland destinations, rural and urban destinations, UNESCO world heritage cities and among coastal regions (Cantabrian Sea, Mediterranean Sea, Canary Islands and Balearic Islands). Location coefficient according to income level To identify destinations specialisation according to tourists’ income level, the location coefficient of destinations has been calculated according to: 𝐿𝑄𝑑𝑖 = 𝑇𝑑𝑖 / 𝑇𝑑 𝑇𝑖/𝑇𝑡 Where LQdi is the location coefficient of municipality d at income level i, Tdi is the number of travellers from income level i travelling to municipality d, Td is the total number of travellers travelling to municipality d, Ti is the number of travellers from income level i at the national level, Tt is the total number of travellers at the national level. This classic indicator, used in economic base theory and geography (Brown & Chung, 2006), compares the share of travellers from a specific group in a destination with their share in the entire country. Values higher than 1 indicate over-representation in a destination, while values below 1 suggest the opposite. Therefore the higher its value, the more concentrated that group is in a destination. The data for this indicator comes directly from the origin-destination matrix of tourist trips and income levels as previously detailed. Bivariate correlations The correlation between the trip matrices for each of the years was analysed to detect the extent to which the matrices for 2020 and 2021 resemble those for the summer of 2019. Significant differences indicate the distortion of tourist trips caused by COVID-19. Multiple regression analysis OLS regression analyses were conducted for each year to determine the factors influencing tourist flows at destinations (dependent variable). The selection of independent variables (Table 2) has based on a literature review of models explaining domestic tourist flows (e.g., Santeramo et al., 2016), domestic tourist flows in Spain (Priego et al., 2015; Alvarez-Diaz et al., 2020) and considering data availability. Socio-economic variables and tourism statistics are from the Spanish National Institute of Statistics. World Heritage Sites were extracted from the UNESCO list. Travel time between origins and destinations was calculated using GIS and TomTom data (TomTom Speed Profiles 2018), for mainland destinations, while air travel time estimates were used for islands destinations. Table 2 outlines the independent variables and their expected sign. Other variables found in the literature review, such as the number of hotel beds at destinations, were omitted due to data limitations (this variable is available for 104 municipalities only). In our study, the greater propensity of tourists for sparsely populated destinations, where contagion risk is lower (Boto- García and Mayor, 2022; Duro et al., 2022) is captured by the population variable. Table 2: Candidate independent variables for OLS regression analyses Variables Description Expected sign (1) Per capita income of destination Net per capita income of the trip destination municipality in euros (2017) (Average value weighted by population for groups of municipalities) (+) Duro, J. A. et al. (2021) Per capita income of origin Net per capita income in euros for census tracts in Madrid (2017) (+) Alvarez-Diaz, M. et al. (2020) Population Number of inhabitants in the destination municipality (2017) (+) Duro, J. A. et al. (2021) Second homes Number of second homes in the destination municipality (2017) (+) Boto-García, D., & Mayor, M. (2022); Falk, M. et al. (2022) Tourist properties Number of tourist properties in the destination municipality (2017) (+) Boto-García, D., & Mayor, M. (2022); Moreno-Luna, L. et al. (2021) World Heritage Cities Dummy variable that takes the value 1 if the destination municipality is a UNESCO world (+) Álvarez-Diaz, M. (2020); Priego, F. J. et al. (2025) heritage site Islands Dummy variable that takes the value 1 if the destination municipality is in the Balearic or Canary Islands (-) Duro, J. A. et al. (2021) Mediterranean Dummy variable that takes the value 1 for destination municipalities on the east and south coasts of mainland Spain (+) Duro, J. A. et al. (2021, 2022) Cantabrian Dummy variable that takes the value 1 for destination municipalities on the north and northwest coasts of mainland Spain (+) Duro, J. A. et al. (2022); Falk, M. et al. (2022) Travel time Average travel time (in minutes) between origin and destination (see text) (-) Álvarez-Diaz, M. et al. (2020); Yang, Y. et al. (2022) (1) Positive sign indicates that an increase in the independent variable produces an increase in the dependent variable and the opposite for negative signs. Source: Authors own creation Table 3 shows the main destination types by population and average per capita income. Rural destinations have a lower average income level than urban destinations and destinations on the Mediterranean coast have a lower average income than for those on the Cantabrian coast. Table 3: Descriptive statistics of per capita income according to major types of tourist destination Number of areas Population Income per person (euros) Mean Std Dev Rural 1,913 6,503,739 10,355.7 2,311.3 Non-Rural 776 32,566,957 10,552.1 2,451.1 World Heritage Cities 41 5,852,387 11,347.2 2,866.4 Islands 117 3,211,549 10,739.8 1,743.2 Mediterranean 183 9,432,195 10,408.2 2,303.8 Cantabrian 121 2,941,054 11,790.6 2,228.7 Total 2,689 39,070,696 10,418.0 2,357.9 Source: Authors own creation THE IMPACT OF THE PANDEMIC ON TOURIST TRIPS BY THE PEOPLE OF MADRID Changes in the number of trips In August 2019, people from Madrid took around 4,250,000 domestic trips (1.39 per person). This figure dropped to 2,986,000 (0.98 trips per person) in August 2020, a 29.7% decline. In the summer 2021, there were 3,510,000 trips (1.15 trips per person), a 17.4% decrease from the pre-pandemic situation(Table 3). The decrease in trip numbers was consistent across the income levels. Table 3: Number of tourists from Madrid and trips per person by income levels in the 2019, 2020 and 2021 Total trips by income levels Q1 Q2 Q3 Q4 Total August-19 1,277,792 1,100,333 1,135,693 736,996 4,250,814 August-20 922,671 776,148 783,547 504,667 2,987,033 August-21 1,053,163 949,908 914,846 592,515 3,510,432 Trips/person by income levels August-19 1.22 1.40 1.55 1.50 1.39 August-20 0.88 0.99 1.07 1.03 0.98 August-21 1.00 1.21 1.25 1.21 1.15 Differences 2020-2019 -27.8% -29.5% -31.0% -31.5% -29.7% 2021-2019 -17.6% -13.7% -19.4% -19.6% -17.4% Source: Authors own creation The pandemic also affected tourist mobility in terms of time/distances travelled. The average travel time decreased between 2019 and 2020 across all income levels, while the proportion of short trips (less than 120 minutes) increased between 2019 and 2020 (Table 4). Table 4: Percentage of trips by travel time (minutes) and income levels. 2019 Q1 Q2 Q3 Q4 Total Short < 120 minutes 29.2% 24.9% 21.3% 19.8% 24.3% Medium 120-240 minutes 53.0% 53.5% 54.3% 52.9% 53.4% Long > 240 minutes 17.9% 21.7% 24.5% 27.3% 22.3% Average time (minutes) 208.6 220.0 228.7 233.8 221.3 2020 Short < 120 minutes 34.2% 28.7% 22.8% 21.0% 27.6% Medium 120-240 minutes 50.5% 52.4% 54.1% 53.0% 52.4% Long > 240 minutes 15.3% 18.8% 23.1% 26.0% 20.1% Average time (minutes) 197.8 211.1 225.2 230.8 214.0 2021 Short < 120 minutes 28.9% 24.0% 20.5% 18.4% 23.6% Medium 120-240 minutes 52.9% 53.5% 54.2% 53.3% 53.4% Long > 240 minutes 18.2% 22.6% 25.4% 28.3% 23.0% Average time (minutes) 209.3 222.2 231.2 237.0 223.2 Source: Authors own creation Changes in the number of trips according to destination Figure 1 shows the distribution of visitors by destination in August 2020 (top) and 2021 (bottom)1. Sphere size represents the total number of visitors, while colours indicate the percentage drop compared to the pre-pandemic situation (August 2019). Main tourist destinations are located in coastal areas, islands, inland rural areas and World Heritage Cities. 1 Non-parametric tests (Wilcoxon and Friedman), were employed to assess the difference before and after the pandemic, given that the annual series data are correlated and do not follow to a normal distribution. Across all cases, including totals and distributions based on destination types, the results consistently indicated significant differences (p value < 0.01) among the three years analyzed. In 2020, most destinations experienced a decline in visitors (red colours), but some, particularly rural municipalities near Madrid and some northern coast municipalities, saw an increase (blue colours). These are less crowded destinations, where tourists probably felt safer. Municipalities along the Catalan coast, including Barcelona, and some of the large coastal destinations and inland cities were among the big losers. In 2021, winning destinations increased, including rural areas near Madrid, the Northern coast, but also in several sectors of the Mediterranean coast, especially Balearic Islands, Costa del Sol (Malaga) and Costa Blanca (Alicante). Recovery was also noted in the Galician coast (northwest) and many municipalities along the Catalan coast. Figure 1. Changes in the number of trips by tourists from Madrid in 2020 (top) and 2021 (bottom). Only including destinations with more than 250 tourists. Source: Authors own creation In the period from 2019 to 2020, urban destinations experienced a greater decline in tourists compared to non-urban destinations (-33.3% and -23.2%, respectively). The Cantabrian coast lost fewer tourists than the overcrowded Mediterranean coast (-22.4% and -31.6%, respectively). Surprisingly, island destinations (Balearic and Canary Islands) recorded a smaller drop in tourists numbers than mainland Spain (-24.0% and -30.0%, respectively), despite air travel requirements. Between August 2021 and August 2019, overall decreases and variations between geographical areas (Table 5) were smaller, suggesting an evolution towards a normal situation in tourist mobility. Table 5. Distribution of tourists from Madrid by destination type. Destination type 2019 2020 2021 Difference 2019-2020 Difference 2019- 2021 Total % Total % Continental Spain 4,055,642 2,838,542 3,339,949 -1,217,100 -30.0 -715,693 -17.6 Islands 195,102 148,371 170,483 -46,731 -24.0 -24,619 -12.6 Urban 2,755,027 1,838,209 2,259,093 -916,818 -33.3% -495,934 -18.0% Non-urban 1,495,791 1,148,735 1,251,343 -347,056 -23.2% -244,448 -16.3% Atlantic coast 532,827 413,440 442,834 -119,387 -22.4 -89,993 -16.9 Mediterranean coast 1,377,259 942,138 1,197,979 -435,121 -31.6 -179,280 -13.0 Total 4,250,744 2,986,913 3,510,432 -1,263,831 -29.7 -740,312 -17.4 Source: Authors own creation Changes in the number of visitors to tourist destinations by tourist income levels. Figures 2 and 3 illustrate destination specialization based on income. In rural areas and the Mediterranean coast, low-income individuals from Madrid are more common, while the Cantabrian coast, the Strait of Gibraltar area, and the Balearic Islands predominantly attract the high-income group. In rural destinations (Figure 3), the drop in the low-income tourists in 2020 was smaller than for high-income tourists. Conversely, on the Mediterranean coast, both income levels experienced similar declines. The Cantabrian coast saw a slightly smaller drop in high-income tourists compared to low-income tourists. The Balearic and Canary Islands stood out, with a much greater decrease in visitors among low-income tourists. Differences between income levels increased in 2021, with higher-income tourists recovering more strongly in most destinations. This is most striking in the case of the coastal destinations, especially the Balearic Islands (Figure 3). Figure 2. Number of tourists from Madrid by destination and location coefficient by the income level of tourists in 2020 Source: Authors own creation Figure 3. Distribution of tourists from Madrid by destination and income level. Source: Authors own creation Bivariate correlation analysis: comparison between tourist trip matrices The correlations between the 2019, 2020 and 2021 trip matrices (Figure 4), show a shallower line slope for the relation of 2019 with 2021 compared to the correlation of 2019 with 2020. This reflects a greater drop in travel in 2020 than in 2021. Destinations below the regression line in the 2019-2020 comparison, such as large cities like Barcelona and Malaga, were particularly affected by the COVID-19 pandemic crisis. Those above the line, such as Torrevieja, Benidorm and Alicante, withstood the first shock of the pandemic better, probably because many people from Madrid have second homes there, providing a greater sense of security than hotels. The correlation coefficients also reveal the intensity of the pandemic’s impact on domestic tourist travel. The lower correlation coefficient value in the 2019-2020 situation (0.77) -21% -22% -26% -26% -17% -13% -18% -18% 0 100000 200000 300000 400000 500000 600000 Low Medium-low Medium-high High Rural 2019 2020 2021 -31% -42% -36% -39% -24% -20% -26% -23% 0 20000 40000 60000 80000 100000 120000 Low Medium-low Medium-high High World Heritage Cities 2019 2020 2021 -32% -35% -28% -31% -18% -4% -17% -13% 0 50000 100000 150000 200000 250000 300000 350000 400000 450000 Low Medium-low Medium-high High Mediterranean -24% -26% -19% -22% -24% -10% -20% -15% 0 20000 40000 60000 80000 100000 120000 140000 160000 180000 Low Medium-low Medium-high High Cantabrian -36% -26% -20% -21% -38% -23% -18% -17% 0 5000 10000 15000 20000 Low Medium-low Medium-high High Canary Islands -38% -25% -23% -13% -23% -4% -1% -2% 0 10000 20000 30000 40000 50000 Low Medium-low Medium-high High Balearic Islands compared to the 2019-2021 correlation (0.93), reflects the disruption to the travel matrix caused by COVID-19. The disruption was more pronounced in 2020 than in 2021, when a clear recovery in domestic tourism is evident. Figure 4: Bivariate correlations between 2019 and 2020 trip matrices (green line) and 2021 (orange line) trip matrices for the top 100 destinations. The black line shows the correlation of the 2019 matrix with itself as a reference. Source: Authors own creation Disaggregating the correlations by income quartile (Table 6) shows that the pandemic’s impact was lower on the low-income level (with a coefficient of determination of 0.83 between 2019 - 2020 travel matrices) and the low-income group saw the strongest recovery in 2021 (coefficient of determination of 0.952 between 2019 – 2021 matrices). Coefficients of determination increase among similar income levels, suggesting that destination choice patterns vary progressively with increasing differences in income. Interestingly, the pandemic exacerbated differences between income levels. In 2019, the coefficients of determination between Q1 and Q2, and Q1 and Q4, were 0.785 and 0.234, respectively, increasing to 0.812 and 0.180 by 2020. Table 6. Coefficients of determination (r2) in the bivariate relationships for the number of visitors by income quartile and year Q1_19 Q2_19 Q3_19 Q4_19 Q1_20 Q2_20 Q3_20 Q4_20 Q1_21 Q2_21 Q3_21 Q4_21 Q1_2019 1 Q2_2019 0.785 1 Q3_2019 0.454 0.828 1 Q4_2019 0.234 0.576 0.867 1 Q1_2020 0.830 0.554 0.29 0.139 1 Q2_2020 0.812 0.732 0.562 0.37 0.87 1 Q3_2020 0.469 0.623 0.754 0.673 0.48 0.789 1 Q4_2020 0.180 0.341 0.598 0.775 0.196 0.449 0.809 1 Q1_2021 0.952 0.743 0.425 0.231 0.809 0.786 0.457 0.192 1 Q2_2021 0.832 0.916 0.715 0.488 0.636 0.79 0.636 0.352 0.879 1 Q3_2021 0.492 0.792 0.91 0.815 0.349 0.627 0.799 0.658 0.533 0.817 1 Q4_2021 0.219 0.499 0.769 0.929 0.147 0.382 0.700 0.848 0.251 0.498 0.838 1 Source: Authors own creation OLS regression analysis OLS multiple regression analyses for summers 2019, 2020 and 2021 (Table 7) revealed a sharp drop of model fit (adjusted R2) from 2019 (0.708) to 2020 (0.589), before partially recovering in 2021 (0.683). This again suggests the pandemic’s strong impact on the tourist flow system, disrupting the previous pattern and partially correcting itself in 2021. These results align with correlation analysis. The trend towards disruption is also evident in fewer significant independent variables and the lower coefficients in the 2020 model compared to 2019. All the independent variables were significant in the three years, except for the population of origin and the two income-level variables. The population of origin is not significant possibly due to similar sizes in grouped census tracts. In terms of the income level variables, the segmentation of destinations into several dummy variables (Mediterranean, Cantabrian, Islands, World Heritage Sites), probably captures the heterogeneity of income differences. The population of the destination and islands variables were not significant in 2020 either. All significant variables have the expected signs. Larger populations with more tourism infrastructure (second homes and tourist properties) and higher attractiveness (Mediterranean and Cantabrian coasts, World Heritage Cities) show positive signs. Conversely, greater travel times and insularity are associated with a reduction in the number of trips. Table 7: Result of OLS analysis. Dependent variable: tourists from Madrid arriving at municipalities in August by year Independent variables 2019 2020 2021 Constant 2033.385*** 1692.295*** 1654.854*** Population of the destination 0.016*** - 0.008*** Second homes 0.744*** 0.612*** 0.632*** Tourist properties 1.177*** 0.764*** 1.437*** World Heritage 1596.153*** 1434.65*** 888.910*** Islands -652.442*** - -620.785*** Mediterranean 1047.391*** 662.619*** 935.811*** Cantabrian 2707.383*** 2366.638*** 2268.224*** Travel time -7.812*** -6.469*** -6.292*** Adjusted R-squared 0.708 0.589 0.683 Akaike Information Criterion (AIC) 47192 46561 46605 No. of observations 2585 2585 2585 Only significant variables are included. Probability * p < 0.1, ** p < 0.05, *** p < 0.001. All VIFs are below 2.5. Source: Authors own creation FINAL REMARKS Conclusions We have used mobile phone data to analyse the changes the pandemic has generated in domestic tourism, focusing on Madrid’s residents. The analysis reveals significant changes in tourist patterns, including a decline in the overall number of trips and a shift towards less crowded destinations, a trend also found by other authors (Duro, J. et al., 2022, Falk, M. et al. 2022). In the summer of 2020, there was a large drop in the number of tourist trips (30%), a figure in line with other studies performed in Spain at the provincial level (Boto-García, D., & Mayor, M. 2022). By destination, the Cantabrian coast, rural areas, and the islands were the least affected destinations, while urban destinations and World Heritage Cities were the “big losers”. Some destinations were particularly affected, such as Barcelona, with very sharp falls in the number of tourists received. In 2021, with the vaccination process underway, there was a recovery in the total number of trips (17% lower than in the summer of 2019). By destination, traditional Mediterranean towns recovered strongly. The islands, especially the Balearic Islands, also recovered among visitors from Madrid with a high income. Income level plays a crucial role in tourist destination choices, with lower-income tourists favouring the Mediterranean coast and higher-income tourists preferring the islands and the Cantabrian coast. However, within these areas, there are destinations with a wide variety of situations in terms of social mix. The composition of the tourists at destinations tends to reflect the income disparities of the origins. This has not previously been measured at this level of spatial detail. The pandemic disrupted the previous tourist pattern but showed signs of recovery in 2021. Correlation analysis between trip matrices and OLS multiple regression confirms that the pandemic disrupted the previous pattern in favour of widespread disorder, although the system had moved to reorganise itself by the summer of 2021. Theoretical Implications The data-driven approach offers significant insights into the impact of the pandemic on tourist flows. While the concept of tourist specialization and the identification of vulnerable social groups within the tourist population emerge as important aspects for policymaking. The findings support the use of mobile phone data and GIS tools to analyse and understand tourism dynamics. Thanks to the high temporal and spatial granularity of mobile phone data, it has been possible to analyse the trips made by people from Madrid in August 2019, 2020 and 2021. This information is not easily available and complements official Spanish statistics such as Familitur or the Hotel Occupancy Survey, used by different authors when analysing domestic tourism and the impact of the pandemic (Álvarez-Diaz, M. et al. 2020, Boto-García, D., & Mayor, M. 2022). Practical Implications The analysis can help policymakers identify regions that have experienced significant losses and target support to those most affected. For example, authorities can design compensatory measures like vouchers or discounts, aimed at providing additional support to those who need it most. Such measures not only encourage tourism, but also foster inclusivity and equitable growth across various communities. The data on the specialisation of tourist markets in certain origins can guide the design of targeted marketing campaigns to specific locations with higher demand. Overall, the outcomes in this study contribute to the development of smart tourism initiatives that optimize resources, enhance the tourism experience, and increase destination competitiveness. Limitations and Future Research The study focuses specifically on Madrid residents, limiting direct generalization to the entire country. However, we must recall that Madrid is the main source of domestic tourists, so its impact has significant repercussions for the rest of the country. Future research should explore the use of mobile phone data to analyse tourist flows and explore the impact of the pandemic in other regions. Data representativeness remains a challenge for big data sources, requiring careful analysis and validation. Nevertheless, mobile phone data stands out as one of the most reliable sources, given the widespread use of mobile devices in contemporary society. Furthermore, our outcomes were consistent with the existing literature. Overall, the study demonstrates the potential of mobile phone data to provide valuable insights into the impact of the pandemic on domestic tourism and inform effective revitalization strategies. Therefore, the development of innovative tools and methodologies to leverage mobile phone data for tourism analysis is encouraged. Acknowledgements The authors would like to thank Kido Dynamics for the data provided and their processing. This work was supported by grants TourMobility (PID2021-127499OB-I00) and DARUMA (PCI2020- 120706-2) funded by MICIU/AEI/10.13039/501100011033 and by “ERDF A way of making Europe”. References Ahas, R., Aasa, A., Mark, Ü., Pae, T., & Kull, A. (2007). Seasonal tourism spaces in Estonia: Case study with mobile positioning data. Tourism Management, 28(3), 898–910 Alvarez‐Diaz, M., D'Hombres, B., Ghisetti, C., & Pontarollo, N. (2020). Analysing domestic tourism flows at the provincial level in Spain by using spatial gravity models. International Journal of Tourism Research, 22(4), 403-415. Arbulú, I., Razumova, M., Rey-Maquieira, J., & Sastre, F. (2021). Can domestic tourism relieve the COVID-19 tourist industry crisis? The case of Spain. Journal of Destination Marketing & Management, 20, Article 100568. Azis, N., Amin, M., Chan, S., & Aprilia, C. (2020). How smart tourism technologies affect tourist destination loyalty. Journal of hospitality and tourism technology, 11(4), 603-625. Backer, E., & Ritchie, B. W. (2017). VFR travel: A viable market for tourism crisis and disaster recovery? International Journal of Tourism Research, 19(4), 400–411. Bae, S. Y., & Chang, P. J. (2021). The effect of coronavirus disease-19 (COVID-19) risk perception on behavioural intention towards ‘untact’tourism in South Korea during the first wave of the pandemic (March 2020). Current Issues in Tourism, 24(7), 1017-1035. Baggio, R., Micera, R., & Del Chiappa, G. (2020). Smart tourism destinations: a critical reflection. Journal of Hospitality and Tourism Technology, 11(3), 407-423. Barros, C., Moya-Gómez, B., and Gutiérrez, J. (2020). Using geotagged photographs and GPS tracks from social networks to analyse visitor behaviour in national parks. Current Issues in Tourism, 23 (10), 1291-1319. Boto-García, D., & Mayor, M. (2022). Domestic tourism and the resilience of hotel demand. Annals of Tourism Research, 93, 103352. Brown, L. A., & Chung, S. Y. (2006). Spatial segregation, segregation indices and the geographical perspective. Population, space and place, 12(2), 125-143. Cerón, M., Palermo, C. M., & Grechi, D. (2021). COVID-19 response models and divergences within the EU: A health dis-union. Statistics, Politics and Policy, 12(2), 219-268. Chantre-Astaiza, A., Fuentes-Moraleda, L., Muñoz, A., & Ramírez, G. (2019). Science Mapping of Tourist Mobility 1980–2019. Technological Advancements in the Collection of the Data for Tourist Traceability. Sustainability, 11(17), 4738. Chu, C. P., & Chou, Y. H. (2021). Using cellular data to analyze the tourists' trajectories for tourism destination attributes: A case study in Hualien, Taiwan. Journal of Transport Geography, 96, 103178. Cui, F., Liu, Y., Chang, Y., Duan, J., & Li, J. (2016). An overview of tourism risk perception. Natural Hazards, 82, 643-658. Duro, J. A., Perez-Laborda, A., Turrion-Prats, J., & Fernández-Fernández, M. (2021). COVID-19 and tourism vulnerability. Tourism Management Perspectives, 38, 100819. Duro, J., Perez-Laborda, A., & Fernandez, M. (2022). Territorial tourism resilience in the COVID- 19 summer. Annals of Tourism Research Empirical Insights, 3(1), 100039. Edsberg Møllgaard, P., Lehmann, S., & Alessandretti, L. (2022). Understanding components of mobility during the COVID-19 pandemic. Philosophical Transactions of the Royal Society A, 380(2214), 20210118. Falk, M., Hagsten, E., & Lin, X. (2022). High regional economic activity repels domestic tourism during summer of pandemic. Current Issues in Tourism, 25(8), 1209-1225. Gao, Y., Sun, D., & Zhang, J. (2021). Study on the impact of the COVID-19 pandemic on the spatial behavior of urban tourists based on commentary big data: A case study of Nanjing, China. ISPRS International Journal of Geo-Information, 10(10), 678. García-Palomares, J. C., Gutiérrez, J., & Mínguez, C. (2015). Identification of tourist hot spots based on social networks: A comparative analysis of European metropolises using photo-sharing services and GIS. Applied Geography, 63, 408-417. Gil-Alana, L. A., & Poza, C. (2022). The impact of COVID-19 on the Spanish tourism sector. Tourism Economics, 28(3), 646-653. Golets, A., Farias, J., Pilati, R., & Costa, H. (2023). COVID-19 pandemic and tourism: The impact of health risk perception and intolerance of uncertainty on travel intentions. Current Psychology, 42(3), 2500-2513. Gössling, S., Scott, D., & Hall, C. M. (2020). Pandemics, tourism and global change: a rapid assessment of COVID-19. Journal of sustainable tourism, 29(1), 1-20. Gretzel, U., Sigala, M., Xiang, Z., & Koo, C. (2015). Smart tourism: foundations and developments. Electronic markets, 25, 179-188. Haneef, S. K. (2017). A model to explore the impact of tourism infrastructure on destination image for effective tourism marketing. University of Salford (United Kingdom). Johnson, A. G., & Samakovlis, I. (2019). A bibliometric analysis of knowledge development in smart tourism research. Journal of Hospitality and Tourism Technology, 10(4), 600-623. Kupfer, J. A., Li, Z., Ning, H., & Huang, X. (2021). Using mobile device data to track the effects of the COVID-19 Pandemic on spatiotemporal patterns of national park visitation. Sustainability, 13(16), 9366. Kuusik, A., Tiru, M., Ahas, R., & Varblane, U. (2011). Innovation in destination marketing: The use of passive mobile positioning for the segmentation of repeat visitors in Estonia. Baltic Journal of Management, 6(3), 378–399. Liu, S., & Zheng, D. (2023). Impacts of tourists’ trust, perception and acceptance of health quick response technology on responsible pandemic travel behaviours. Journal of Hospitality and Tourism Technology, 14(2), 278-294. Mehraliyev, F., Choi, Y., & Köseoglu, M. A. (2019). Progress on smart tourism research. Journal of Hospitality and Tourism Technology, 10(4), 522-538. Moreno-Luna, L., Robina-Ramírez, R., Sánchez, M. S. O., & Castro-Serrano, J. (2021). Tourism and sustainability in times of COVID-19: The case of Spain. International Journal of Environmental Research and Public Health, 18(4), 1859. Neuburger, L., & Egger, R. (2021). Travel risk perception and travel behaviour during the COVID- 19 pandemic 2020: A case study of the DACH region. Current issues in tourism, 24(7), 1003-1016. Nilbe, K., Ahas, R., & Silm, S. (2014). Evaluating the travel distances of events visitors and regular visitors using mobile positioning data: The case of Estonia. Journal of Urban Technology, 21(2), 91–107. Pérez-Arnal, R., Conesa, D., Alvarez-Napagao, S., Suzumura, T., Català, M., Alvarez-Lacalle, E., & Garcia-Gasulla, D. (2021). Comparative analysis of geolocation information through mobile- devices under different Covid-19 mobility restriction patterns in Spain. ISPRS International Journal of Geo-Information, 10(2), 73. Priego, F. J., Rosselló, J., & Santana-Gallego, M. (2015). The impact of climate change on domestic tourism: a gravity model for Spain. Regional environmental change, 15(2), 291-300. Qian, C., Li, W., Duan, Z., Yang, D., & Ran, B. (2021). Using mobile phone data to determine spatial correlations between tourism facilities. Journal of Transport Geography, 92, 103018. Raun, J., Ahas, R., & Tiru, M. (2016). Measuring tourism destinations using mobile tracking data. Tourism Management, 57, 202–212. Ren, M., Park, S., Xu, Y., Huang, X., Zou, L., Wong, M. S., & Koh, S. Y. (2022). Impact of the COVID- 19 pandemic on travel behavior: A case study of domestic inbound travelers in Jeju, Korea. Tourism Management, 92, Article 104533. Santeramo, F. G., & Morelli, M. (2016). Modelling tourism flows through gravity models: A quantile regression approach. Current Issues in Tourism, 19(11), 1077-1083. Shoval, N., & Ahas, R. (2016). The use of tracking technologies in tourism research: the first decade. Tourism Geographies, 18(5), 587-606. Sobolevsky, S., Sitko, I., Des Combes, R. T., Hawelka, B., Arias, J. M., & Ratti, C. (2014). Money on the move: Big data of bank card transactions as the new proxy for human mobility patterns and regional delineation. the case of residents and foreign visitors in Spain. In 2014 IEEE international congress on big data (pp. 136-143). IEEE. Timur, B., Oğuz, Y. E., & Yilmaz, V. (2023). Consumer behavior of mobile food ordering app users during COVID-19: dining attitudes, e-satisfaction, perceived risk, and continuance intention. Journal of Hospitality and Tourism Technology, 14(3), 460-475. Tiru, M., Kuusik, A., Lamp, M.-L., & Ahas, R. (2010). Lbs in marketing and tourism management: Measuring destination loyalty with mobile positioning data. Journal of Location Based Services, 4(2), 120–140. Versichele, M., Neutens, T., Delafontaine, M., & Van de Weghe, N. (2012). The use of Bluetooth for analysing spatiotemporal dynamics of human movement at mass events: A case study of the Ghent Festivities. Applied Geography, 32(2), 208-220. Viana-Lora, A., Domenech, A., & Gutiérrez, A. (2023). COVID-19 and tourist mobility at destinations: a literature review and emerging research agenda. Journal of Tourism Futures, 9(1), 21-34. Wang, L., Wu, X., & He, Y. (2021). Nanjing’s Intracity Tourism Flow Network Using Cellular Signaling Data: A Comparative Analysis of Residents and Non-Local Tourists. ISPRS International Journal of Geo-Information, 10(10), 674. Woo-Hyuk Kim, Eunhye (Olivia) Park and Bongsug (Kevin) Chae (2023). Understanding the impact of COVID-19 on consumer mobility and recovery from a distance perspective: a mobile phone data application. Journal of Hospitality and Tourism Technology, Vol. ahead-of-print No. ahead-of-print Xu, Y., Xue, J., Park, S., & Yue, Y. (2021). Towards a multidimensional view of tourist mobility patterns in cities: A mobile phone data perspective. Computers, Environment and urban systems, 86, 101593. Yang, Y., Zhang, L.,Wu, L., & Li, Z. (2022). Does distance still matter? Moderating effects of distance measures on the relationship between pandemic severity and bilateral tourism demand. Journal of Travel Research. Yu, L., Zhao, P., Tang, J., & Pang, L. (2023). Changes in tourist mobility after COVID-19 outbreaks. Annals of Tourism Research, 103522. Zielinski, S., & Botero, C. M. (2020). Beach tourism in times of COVID-19 pandemic: critical issues, knowledge gaps and research opportunities. International Journal of Environmental Research and Public Health, 17(19), 7288.