Fig 1.
Impact of age and gender on selected parameters of customers’ spending behavior.
A: Average number of transactions per year. B: Average value of a single purchase. Insets: Logarithmic distributions for both genders and different age groups.
Fig 2.
Impact of age and gender on customers’ spending diversity.
A: Average spending diversity against age for men and women. B: Normalized Herfindahl-Hirschman Index of visited business categories for the customers with different levels of spending activity.
Fig 3.
Impact of age and gender on customer mobility.
A: Frequency of distant travels. B: Average distance to home of local transactions. Insets: Logarithmic distributions for both genders and different age groups.
Fig 4.
Superlinear scaling of total spending activity with city size for the Conurbation level.
Total spending activity is defined as the cumulative number of transactions made by city residents. Scaling exponent: 1.045, confidence interval: [1.03,1.06], p-value: 4 ⋅ 10−205, R2 = 99.0%.
Fig 5.
Superlinear scaling of total spending activity with city size.
A: For the level of Large Urban Zones. Scaling exponent: 1.052, CI: [1.00,1.10], p-value: 5 ⋅ 10−23, R2 = 98.90%. B: For the level of Functional Urban Areas. Scaling exponent: 1.044, CI: [1.00,1.08], p-value: 1 ⋅ 10−37, R2 = 98.71%. Total spending activity is defined as the cumulative number of transactions made by city residents.
Table 1.
Scaling of customers’ activity with CON size for different business categories.
Fig 6.
Deviations of spending parameters from their respective scaling trends with city size, for the cities defined at the level Large Urban Zones (LUZ), Functional Urban Areas (FUA), and Conurbations (CON).
Colors indicate three clusters of cities obtained based on the k-mean clustering, in accordance with Fig 7.
Fig 7.
Classification of Spanish cities into three categories based on the spending behavior of their residents.
Classification was performed separately for the three different city definition levels–Large Urban Zones (LUZ), Functional Urban Areas (FUA), and Conurbations (CON). Clustering into two categories can be, to large extent, recreated by merging clusters B and C into one (the consistency between the two- and three-cluster cases was quantified as high as 100% for LUZ, 98% for FUA, and 97% for CON).
Fig 8.
Deviations of standard socioeconomic statistics within the three clusters of Spanish cities at the level of LUZ, and FUA.
Unemployment, Disposal Annual Income of Households, and Gross Domestic Product were available for both city levels. Total Expenditure of Households, from the Household Budget Survey, was available only for the Autonomous Communities and reassigned to the corresponding LUZ. The data comes for the Urban Audit survey of Eurostat [42] and the Spanish National Statistics Institute [45]. Socioeconomic parameters were quantified as log residuals from their respective scaling trends with city size. Colors correspond to the three clusters of cities obtained based on the spending behavior of city residents, as presented in Fig 7.
Fig 9.
Correspondence of the received clusters of Spanish cities with the attractiveness of these cities.
Attractiveness is calculated based on the spending activity of foreign visitors. Presented deviations are quantified as log residuals from their respective scaling trends with city size at the levels of LUZ, FUA and CON. Colors correspond to the three clusters of cities obtained based on the spending behavior of city residents, as presented in Fig 7.