Unraveling the association between socioeconomic diversity and consumer price index in a tourism country

The value of diversity in modern society is widely acknowledged, with empirical evidence linking it to increased productivity and innovation. However, the quantitative relationship between micro-level diversity and macro-level economic indicators like the CPI remains largely unexplored due to data limitations. This paper addresses this gap by leveraging high-resolution mobile phone data (call detail records or CDRs) to quantify the relationship between socioeconomic diversity and CPI in Andorra, a country heavily reliant on tourism and exhibiting high tourist diversity. The study uses CDR data to extract measures of socioeconomic diversity, focusing on nationality and income, at the cell tower level. These diversity measures are then correlated with various CPI measures, including the general CPI and sector-specific CPIs for tourism and resident-related sectors. The high spatial and temporal resolution of mobile phone data offers a unique opportunity to monitor CPI at a granular level, potentially improving the timeliness and accuracy of existing macroeconomic indicators, particularly valuable in developing countries where reliable economic data is often scarce. This study contributes to two growing research areas: linking diversity to economic outcomes, and using large-scale data for macroeconomic monitoring.

Existing economic literature presents varied perspectives on the relationship between diversity and economic growth. Some studies suggest a positive correlation between diversity and economic development, citing benefits such as increased productivity and innovation. Conversely, other research highlights potential negative impacts due to resource allocation issues or social frictions among different groups. Prior studies have also explored the relationship between ethnic diversity and financial market stability, or the impact of workforce diversity on corporate profits. Studies have also attempted to predict macroeconomic indicators such as inflation and CPI using financial market data. However, there is a lack of research quantifying the relationship between micro-level diversity measures and macro-level economic indicators like the CPI. The scarcity of data on both micro-behaviors and macroeconomic indicators has hindered this research. This study seeks to bridge this gap by employing large-scale mobile phone data to analyze the relationship between diversity and CPI.

This study uses call detail records (CDRs) from Andorra's sole mobile carrier, covering the period from July 2014 to August 2016. CDRs offer high spatial and temporal resolution data on mobile phone activity, including location (cell tower), timestamp, SIM card registry country, and phone characteristics (brand, model). The researchers use Voronoi tessellation to approximate cell tower coverage areas. Cell towers are manually categorized into eight types of Points of Interest (POIs): wellness, leisure, shop, gastronomic, nature, event, culture, and others. Diversity measures are calculated using Shannon entropy, focusing on nationality (10 distinct nationalities) and approximated disposable income (14 price categories based on phone prices). Shannon entropy captures the evenness of distribution across categories (nationality and income). The diversity measures are computed daily for each cell tower and then aggregated to the POI level. Monthly CPI data for the general CPI and various sectors (tourism and resident-related) is collected from the Andorran Government. The change in CPI relative to the previous month is calculated. The study explores the correlation between the diversity measures and CPI changes using Pearson correlation. Elastic net regression is employed to select the most important diversity measures for predicting each CPI category, providing a parsimonious model. Finally, the selected diversity measures are used to construct daily and spatial maps of the CPI, allowing for high-resolution nowcasting of CPI at the cell tower level.

The study reveals strong correlations between diversity measures and CPI in Andorra. Income diversity at leisure and nature POIs exhibits a particularly strong positive correlation with general CPI (0.805 and 0.775 respectively, p<0.001). Nationality diversity at the country level shows a strong negative correlation with the general CPI (-0.650, p<0.001) and some sector-specific CPIs (e.g., food, drinks, and tobacco, clothes and shoes). Conversely, income diversity shows a positive correlation with general CPI and several resident-related CPIs (transport, resident services). Elastic net regression identifies a smaller subset of diversity measures that are highly predictive of different CPI categories. For example, only four diversity measures are sufficient to reasonably predict the CPI for hotels and restaurants. For general CPI, the model including eight diversity measures achieves an R² of 0.74, with income diversity measures contributing positively and nationality diversity negatively. Daily nowcasting of CPI using the selected variables accurately captures the temporal variations in CPI across different sectors. Spatial mapping reveals regional variations in CPI within Andorra, highlighting the limitations of relying solely on national-level CPI figures. The study demonstrates the potential of using mobile phone data to generate high-resolution daily and spatial CPI maps.

The findings suggest a strong association between socioeconomic diversity and CPI in Andorra, providing empirical evidence for the link between social structures and economic indicators. The positive association between income diversity and CPI may reflect increased consumer spending due to the presence of high-income tourists. The negative correlation between nationality diversity and certain CPI categories could indicate that greater diversity in tourist origins leads to competitive pricing pressures, resulting in lower prices in some sectors. The ability to nowcast CPI at a daily and regional level using mobile phone data offers significant advantages for policymakers. This high-resolution information can enable timely interventions and more effective economic policy adjustments. The results support the potential of utilizing large-scale behavioral data for macroeconomic monitoring, offering valuable insights that traditional statistics might miss.

This study demonstrates a strong association between socioeconomic diversity measures derived from mobile phone data and CPI in Andorra, a tourism-dependent country. The findings highlight the potential of using such data for high-resolution nowcasting and mapping of CPI, providing valuable insights for policymakers. The study's limitations include its reliance on observational data which prevents causal inference. Future research should investigate the causal mechanisms underlying the observed associations and explore the generalizability of these findings to other countries with different economic structures and levels of tourism dependence.

The study's primary limitation is its inability to establish causality between diversity and CPI. The correlational analysis demonstrates a strong association but cannot definitively determine a causal relationship. The study focuses solely on Andorra, a small tourism-dependent country, limiting the generalizability of the results. The use of phone prices as a proxy for income might introduce some measurement error. Future research should address these limitations through causal inference methods and cross-country comparisons.