MTA Use Case


This vignette describes a typical use case of the MTA package with a concrete example: the analysis of income inequalities in the Greater Paris Metropolitan Area. This example plays all the functions of the MTA package with appropriate visualizations (plots, maps).


1. Introducing case-study and MTA

1.1 Multiscalar Territorial Analysis for Policy Study

Beginning in the MTA conceptual framework, several issues must be considered: the study area, the territorial hierarchy and the selected indicator.

This vignette proposes some concrete outputs for improving the knowledge on income inequalities and proposing solutions for a better geographical distribution of wealth in the MGP area. Relevant statistics will be computed using MTA functionalities and plotted on graphics and maps.

1.2 Dataset and complementary Packages

The example dataset, GrandParisMetropole includes all the required information to create

Two additional packages are used: mapsf for thematic mapping purposes and ineq for computing inequality indexes and Lorenz Curve Plot.

# load packages
library(sf)
library(MTA)
library(mapsf)
library(ineq)

# load dataset
com <- st_read(system.file("metroparis.gpkg", package = "MTA"),
               layer = "com", quiet = TRUE)
ept <- st_read(system.file("metroparis.gpkg", package = "MTA"), 
               layer = "ept", quiet = TRUE)

# set row names to municipalities names
row.names(com) <- com$LIBCOM

1.3 Context Maps

Context maps highlight the territorial organization of the study area and provide some first insights regarding the spatial patterns introduced by the indicator used for the analysis.

1.3.1 Study Area

The 150 municipalities of the MGP belongs to 12 intermediate territorial divisions: the Établissements Publics Territoriaux (EPT). This territorial zoning follows approximately the delineation of the départements : Seine-Saint-Denis (in blue on the map below); Paris (in purple); Hauts-de-Seine (in green) and _Val-de-Marine (in orange).

# label / colors management
LIBEPT <- c("Paris", "Est Ensemble", "Grand-Paris Est", 
            "Territoire des aeroports", "Plaine Commune", "Boucle Nord 92", 
            "La Defense","Grand Paris Sud Ouest", "Sud Hauts-de-Seine", 
            "Val de Bievres - Seine Amond - Grand Orly",
            "Plaine Centrale - Haut Val-de-Marne - Plateau Briard",
            "Association des Communes de l'Est Parisien")

# colors
cols <- c("#cfcfcf", # Grey(Paris)
          "#92C8E0", "#7BB6D3", "#64A4C5", "#458DB3", # Blues (dept 93)
          "#A6CC99", "#8CBB80", "#71A966", "#4E9345", # Greens (dept 92)
          "#F38F84", "#EF6860", "#EA3531") # Reds (dept 94)

colEpt <- data.frame(LIBEPT, cols)

# zoning
mf_map(x = com, var="LIBEPT", type = "typo",
       val_order = LIBEPT, pal = cols, lwd = 0.2, border = "white",
       leg_pos = "left", leg_title = "EPT of belonging")

mf_map(ept, col = NA, border = "black", add = TRUE)

# layout 
mf_layout(title = "Territorial Zoning of the MGP",
          credits = paste0("Sources : GEOFLA® 2015 v2.1, Apur, impots.gouv.fr",
                           "\nRonan Ysebaert, RIATE, 2021"))

1.3.2 Map numerator and denominator

The numerator (number of tax households) and the denominator (total amount of income tax references) building the reference ratio (average income per tax households) are firstly plotted.

# layout
opar <- par(mfrow = c(1, 2))

# numerator map
com$INCM <- com$INC / 1000000
mf_map(com,  col = "peachpuff",  border = NA)
mf_map(ept,  col = NA,  border = "black", add = TRUE)

mf_map(x = com, var = "INCM", type = "prop",
       col =  "#F6533A", inches = 0.15, border = "white",
       leg_pos = "topleft", leg_val_rnd =  0,
       leg_title = "Amount of income taxe reference\n(millions of euros)")

# layout
mf_layout(title = "Numerator - Amount of income tax reference",
          credits = paste0("Sources : GEOFLA® 2015 v2.1, Apur, impots.gouv.fr",
                           "\nRonan Ysebaert, RIATE, 2021"), 
          arrow = FALSE)


# denominator map
mf_map(com,  col = "peachpuff",  border = NA)
mf_map(ept,  col = NA,  border = "black", add = TRUE)

mf_map(x = com, var = "TH", type = "prop",
       col =  "#515FAA", inches = 0.15, border = "white",
       leg_pos = "topleft", leg_val_rnd =  -2,
       leg_title = "Number of tax households", add = TRUE)

# layout
mf_layout(title = "Denominator - Tax households",
          credits = paste0("Sources : GEOFLA® 2015 v2.1, Apur, impots.gouv.fr",
                           "\nRonan Ysebaert, RIATE, 2021"), 
          arrow = FALSE)
par(opar)

Without surprise, the highest amounts of tax households and income are located in the central area of the MGP (Parisian Arrondissements). That being said, these two maps suggest an unequal distribution of income when looking at the distribution of population in Paris suburbs.

1.3.3 Ratio

# ratio
com$ratio <- com$INC / com$TH

# ratio map
mf_map(x = com, var = "ratio", type = "choro",
       breaks = c(min(com$ratio, na.rm = TRUE), 
                  20000, 30000, 40000, 50000, 60000, 
                  max(com$ratio, na.rm = TRUE)),
       pal = c("#FCDACA", "#F6A599", "#F07168", "#E92B28", "#C70003", "#7C000C"),
       border = "white", lwd = 0.2, leg_pos = "topleft", leg_val_rnd = 0,
       leg_title = paste0("Average amount of income tax",
                          "\nreference per households\n(in euros)")) 

# EPT borders
mf_map(ept, col = NA, border = "black", add = TRUE)

# Label min and max
mf_label(x = com[which.min(com$ratio),], var = "LIBCOM", cex = 0.6, font = 2,
         halo = TRUE)
mf_label(x = com[which.max(com$ratio),], var = "LIBCOM", cex = 0.6, font = 2,
         halo = TRUE)

# layout 
mf_layout(title = "Ratio - Income per tax households, 2013",
          credits = paste0("Sources : GEOFLA® 2015 v2.1, Apur, impots.gouv.fr",
                           "\nRonan Ysebaert, RIATE, 2021"), 
          arrow = FALSE)

The MGP area is characterized by high income inequalities. For the 150 municipalities of this area, the values extend from 14 730 (La Courneuve) to 96 310 euros (Paris_7th Arrondissement). 53 municipalities of the MGP area (35 % of the municipalities) are below the French average, i.e. 25 660 euros. The lagging households are mainly concentrated into the north part of MGP area. Highest values are concentrated in the Western part of Paris and its suburbs.

1.4 Introducing the MTA Functions

1.4.1 General, territorial and spatial deviations

MTA package introduces three contexts to monitor territorial inequalities: the general, the territorial and the spatial deviations.

The global deviation is dedicated to the analysis of inequalities using a value of reference. In this example the global deviation refers to the inequalities existing between each municipality in regard to the whole MGP average.

The territorial deviation consists in measuring the inequalities existing for each basic territorial unit as regards to an intermediate territorial level of reference. In this case-study, it will be the EPT of belonging of each municipality. It implies to include beforehand in the input dataset a territorial level of belong for each territorial unit.

The spatial deviation is a measure of inequalities taking into account the neighborhood as a reference. It allows to measure the deviation existing between basic territorial units using three possible parameters : the territorial contiguity (order 1, 2 or n), the spatial neighborhood (territorial units located at less than X kilometers) or functional distances (territorial units located at less than Y minutes by road for instance). In MTA, territorial contiguity and spatial neighborhood measures are directly calculated. Functional distances must be uploaded separately through a dataframe structured with the following fields : id1, id2, distance measure. For this case-study, the contiguity criteria (order 1) will be used for the calculation of the spatial deviation. It gives a good proxy of proximity relationships according to the size and the homogeneity of the municipalities of the MGP area. Other measures could be also adapted, such as time-distances by road (municipalities located at less than 15 minutes by car) or by public transportation.

1.4.2 Relative and absolute deviations

In MTA, two methods are implemented to measure statistical differences to a given context of reference: a relative deviation and an absolute deviation. In MTA, each indicator is considered as a ratio defined by a numerator (GDP for instance) divided by a denominator (population for instance).

The relative deviation states the position of each region as regards to a reference context. It is based on the following calculation: Relative deviation (Region i) = 100 * ((Numerator(Region i)/Denominator(Region i)) /(reference ratio)) Territorial units characterized by a context of reference below index 100 are under the average of a given reference context, and reciprocally.

The absolute deviation specifies which process of redistribution should be realized in absolute terms to achieve a perfect equilibrium for the ratio of reference in the global, the territorial or the spatial context. It is calculated as below: Absolute deviation (Region i) = Numerator (Region i) - (reference ratio * denominator (Region i)) In concrete terms, it examines how much amount of the numerator should be moved in order to reach equidistribution, for each territorial unit, taking into account as a reference the selected deviation context value. More generally, absolute deviation must be considered as a statistical tool to discuss on the amplitude of existing territorial inequalities. It is obvious that reaching a perfect equilibrium between territorial units is highly sensitive and is scarcely a policy objective itself. It is nevertheless interesting to consider the amount of money or population affected by the inequality to have in hand a concrete material for leading discussions on this delicate spatial planning issue.

1.4.3 Cartography of relative and absolute deviations

In this vignette, color palette proposed for displaying on map the relative deviations are the ones suggested by the HyperAtlas tool (blue palette = under the average; red palette = above the average). But other diverging palettes (green/red, etc.) could be also be used for displaying MTA results on maps.

Absolute deviations are highlighted using proportional circles on maps. It examines which amount of the numerator should be moved to the poorest municipalities to reach equidistribution. Circles displayed in a red means that the territorial unit have to contribute a given amount of numerator to achieve the equilibrium in a given context; and reciprocally blue circles means that the territorial unit have to receive frmo the others.

It is quite easy to reproduce or adapt these maps using the functions proposed by the mapsf package.

1.4.4 Synthesis

The analysis of the three deviations provides generally a lot of information. The synthesis functions bidev and mst helps to summarize values of the global, territorial and spatial deviations and allow to answer to some basic and interesting questions:

Moreover, some additional functions are implemented to ease the mapping (map_bidev and map_mst) or proposing appropriate plots (plot_bidev and plot_mst) for visualizing these results.


2. Relative deviations

This section aims at describing how computing MTA relative deviations and creating appropriate maps and plots.

2.1 Global deviation

First of all, each territorial unit is compared to the overall study area average (Métropole du Grand Paris).

The code below takes in entry the numerator (INC) and the denominator (TH) of the com object and returns the global deviation indicators using gdev() function. The relative deviation (type = "rel") is computed. This indicator is afterwards mapped.

# general relative deviation
com$gdevrel <- gdev(x = com, 
                    var1 = "INC", 
                    var2 = "TH", 
                    type = "rel")

# Colors for deviations
devpal <-  c("#4575B4", "#91BFDB", "#E0F3F8", "#FEE090", "#FC8D59", "#D73027")

# Global deviation mapping
mf_map(x = com, var = "gdevrel", type = "choro",
       breaks = c(min(com$gdevrel, na.rm = TRUE), 67, 91, 100, 125, 150,
                  max(com$gdevrel, na.rm = TRUE)),
       pal = devpal, border = "white", lwd = 0.2, 
       leg_pos = "topleft", leg_val_rnd = 0,
       leg_title = paste0("Deviation to the global context",
                          "\n(100 = Metropole du Grand Paris average)")) 

# Plot EPT layer
mf_map(ept, col = NA, border = "#1A1A19", lwd = 1, add = TRUE)

# layout 
mf_layout(title = "Global deviation - Tax income per households",
          credits = paste0("Sources : GEOFLA® 2015 v2.1, Apur, impots.gouv.fr",
                           "\nRonan Ysebaert, RIATE, 2021"), 
          arrow = FALSE)

The resulting map highlights strong inequalities: First, it is intersting to know that all the municipalities of the EPT of Plaine Commune, Territoire des Aéroports and Est Ensemble are below the average of the MGP area (33 501 euros per households). For the territories of Val de Bièvres and Grand-Paris Est, only three municipalities are above the average of the study area. Reversely, all the municipalities of the Grand-Paris-Sud-Ouest EPT are largely above the average of the MGP area. For the other EPT the situation is mixed, depending of the municipalities.

The ineq package proposes some functions to depict global inequalities existing in a study area, such as the Lorenz-curve. The Lc function takes in entry the numerator and the denominator and returns a Lorenz Curve plot; inequality indexes take in entry the ratio (numerator / denominator) and returns econometric indexes of inequality.

The code below add some additional graphical parameters to ease the interpretation of this plot.

library(ineq)
#  Concentration of X as regards to concentration of Y
Lc.p <- Lc(com$INC, com$TH)
Lp <- data.frame(cumX = 100 * Lc.p$L, cumY = 100 * Lc.p$p)

# Plot concentrations
opar <- par(mar = c(4,4,1.2,4), xaxs = "i", yaxs = "i", pty = "s")
plot(Lp$cumY, Lp$cumX,  type = "l", col = "red", lwd = 2,
     panel.first = grid(10,10), xlab = "Households (cumulative percentage)",
     ylab = "Income (cumulative percentage)", 
     cex.axis = 0.8, cex.lab = 0.9, ylim = c(0,100), xlim = c(0,100))  

lines(c(0,100), c(0,100), lwd = 2)

# Ease plot reading
xy1 <- Lp[which.min(abs(50 - Lp$cumX)),]
xy2 <- Lp[which.min(abs(50 - Lp$cumY)),]

xy <- rbind(xy1, xy2)

points(y = xy[,"cumX"], 
       x = xy[,"cumY"], 
       pch = 21,  
       cex = 1.5,
       bg = "red")

text(y = xy[,"cumX"], 
     x = xy[,"cumY"],
     label = paste(round(xy[,"cumX"],0), round(xy[,"cumY"],0), sep = " , "),  
     pos = 2,
     cex = 0.6)
par(opar)

The curve depicts on its horizontal axis a defined population – e.g., all households – broken down into deciles and ordered from left to right on the horizontal axis (from the lower tax income per household to the higher). On the vertical axis of the Lorenz curve is shown the cumulative percentage of tax income.

The reading of the plot reveals these following configurations :

2.2 Territorial deviation

The territorial deviation is now calculated to analyze the position of each municipality as regards to a territorial level of reference: the (EPT) of belonging in this case-study.

The tdev function takes in entry the numerator (INC) and the denominator (TH) of the com object. The territorial level of belonging in specified with the key argument. The territorial relative deviation is returned by the function.

# Territorial relative deviation calculation
com$tdevrel <- tdev(x = com, 
                    var1 = "INC", 
                    var2 = "TH", 
                    type = "rel",
                    key = "LIBEPT")

# Cartography
# Territorial deviation mapping
mf_map(x = com, var = "tdevrel", type = "choro",
       breaks = c(min(com$tdevrel, na.rm = TRUE), 67, 91, 100, 125, 150,
                  max(com$tdevrel, na.rm = TRUE)),
       pal = devpal, border = "white", lwd = 0.2, 
       leg_pos = "topleft", leg_val_rnd = 0,
       leg_title = paste0("Deviation to the territorial context",
                          "\n(100 = EPT average)")) 

# Plot EPT layer
mf_map(ept, col = NA, border = "#1A1A19", lwd = 1, add = TRUE)

# Labels to ease comment location
mf_label(x = com[com$LIBCOM %in% c("Le Raincy", "Rungis", "Sceaux",
                                   "Marnes-la-Coquette") ,],
         var = "LIBCOM", cex = 0.6, font = 2, halo = TRUE)

# layout 
mf_layout(title = "Territorial deviation - Tax income per households, 2013",
          credits = paste0("Sources : GEOFLA® 2015 v2.1, Apur, impots.gouv.fr",
                           "\nRonan Ysebaert, RIATE, 2021"), 
          arrow = FALSE)

This map highlights existing inequalities in each EPT: The strongest differences in relative terms are located in Paris (opposition between the eastern part and the western part of this EPT) and in the Plaine centrale - Haut Val de Marne EPT (opposition between the poorest municipalities located near Paris and the ones located in the periphery). Globally, the richest and the poorest EPT ( Grand Paris Sud Ouest / Plaine Commune and Territoires des Aéroports du Nord Ouest) appear relatively homogeneous statistically. In other EPT, one municipality appears largely above the average of their EPT of belonging. It is namely the case in Grand-Paris-Sud-Ouest (Marnes-la-Coquette), Sud Hauts-de-Seine (Sceaux), Grand Orly Seine Biève (Rungis) or Grand Paris Grand Est (Le Raincy)

Another way to explore characteristics of the territorial deviation consists in analyzing the statistical dispersion (general deviation) by intermediate level (EPT in this case). The best suited graphical representation for this is certainly a boxplot.

The code below takes in entry the general deviation calculated above and the intermediate levels included in the input dataset. It returns a boxplot displaying the statistical parameters (median, mean, 1st and 3rd quartiles, range, minimum and maximum, extraordinary values) allowing to observe the statistical dispersion existing for each intermediate zoning. To ease the interpretation and the synthesis of the plot, boxplots are ordered by the average of each territorial level. Moreover, the width of the bars are proportional to the number of territorial units included in each intermediate zoning.

opar <- par(mar = c(4, 4, 1.2, 2))

# Drop geometries
df <- st_set_geometry(com, NULL)

# Reorder EPT according to gdev value
df$EPT <- with(df, reorder(EPT, gdevrel, mean, na.rm = TRUE))

# Colors management
col <- aggregate(x = df[,"gdevrel"], by = list(LIBEPT = df$LIBEPT),
                 FUN = mean)
col <- merge(col, colEpt, by = "LIBEPT")
col <- col[order(col$x),]
cols <- as.vector(col$cols)

# Drop inexisting levels
df <- droplevels(df)

# Boxplot
bp <- boxplot(df$gdevrel ~ df$EPT, col = cols, ylab = "Global deviation",
              xlab = "Territorial deviation", cex.lab = 0.9,
              varwidth = TRUE, range = 1.5,  outline = TRUE,  las = 1) 

# Horizontal Ablines
abline (h = seq(40, 300, 10), col = "#00000060", lwd = 0.5, lty = 3)

# Plot mean values
xi<- tapply(df$gdevrel, df$EPT, mean, na.rm = TRUE)
points(xi, col = "#7C0000", pch = 19)

# Legend for the boxplot
legend("topleft", legend = rev(as.vector(col$LIBEPT)), pch = 15,
       col = rev(as.vector(col$cols)), cex = 0.8, pt.cex = 1.5)
par(opar)

This plot highlights the statistical dispersion existing in each EPT. It confirms globally that wealthier the EPT is, larger the statistical differences between the poorest and the wealthiest territorial units are. In that way, Plaine Commune and Territoire des Aéroports (T6 and T7) appear quite homogeneous in a lagging situation. On the opposite, important differences a revealed in Paris (minimum = 71 and maximum = 290) or in La Defense and Grand Paris Sud Ouest.

The boxplot highlights also outliers (dots out of the box) in each territorial context. Most EPT are concerned: wealthiest EPT (Paris, La Défense, Grand Paris Sud Ouest and ACEP, generally characterized by high outliers) and EPT in less favorable situation (Est Ensemble, Grand Paris Est and Sud-Hauts-de-Seine, which include municipalities with include both low and high outliers.

2.3 Spatial deviation

The spatial deviation is calculated to position territorial units in a neighborhood context. Several criteria may be considered (geographical distance, time-distance). In this example, each municipality will be compared to the average of contiguous municipalities (contiguity order 1).

The sdev function takes in entry the numerator (INC) and the denominator (TH) of the com object. The contiguity order 1 is set in the order argument.

# Spatial relative deviation calculation
com$sdevrel <- sdev(x = com, 
                    xid = "DEPCOM", 
                    var1 = "INC", 
                    var2 = "TH",
                    order = 1,
                    type = "rel")

# Cartography
# Territorial deviation (relative and absolute) cartography
mf_map(x = com, var = "sdevrel", type = "choro",
       breaks = c(min(com$sdevrel, na.rm = TRUE), 67, 91, 100, 125, 150,
                  max(com$sdevrel, na.rm = TRUE)),
       pal = devpal, border = "white", lwd = 0.2, 
       leg_pos = "topleft", leg_val_rnd = 0,
       leg_title = paste0("Deviation to the spatial context",
                          "\n(100 = average of the contiguous",
                          " territorial units - order 1)")) 

# Plot EPT
mf_map(ept, col = NA, border = "#1A1A19",lwd = 1, add = T)

# Labels to ease comment location
mf_label(x = com[com$LIBCOM %in% c("Le Raincy", "Vaucresson", "Sceaux", "Bagneux",
                                   "Marnes-la-Coquette", "Saint-Maur-des-Fosses",
                                   "Puteaux", "Saint-Ouen", "Clichy-sous-Bois", 
                                   "Clichy"),],
         var = "LIBCOM", cex = 0.6, font = 2, halo = TRUE, overlap = FALSE)

# layout 
mf_layout(title = "Spatial deviation - Tax income per households, 2013",
          credits = paste0("Sources : GEOFLA® 2015 v2.1, Apur, impots.gouv.fr",
                           "\nRonan Ysebaert, RIATE, 2021"), 
          arrow = FALSE)

This map highlights local discontinuities existing in the study area. Important local statistical gaps appear in several areas: Saint-Mandé and Neuilly-sur-Seine are characterized by the highest score as regards to their respective neighbors (indexes 173 and 156).Some local "bastions" are revealed in Paris suburbs, such as Le Raincy (index 153), Sceaux (150), Vaucresson (143), Marnes-la-Coquette (142) or Saint-Maur-des-Fossés (140).

On the reverse situation, lower index are observed at Clichy-sous-Bois. Clichy, Puteaux, Saint-Ouen and Bagneux. Their average income per household is around 30-40 % below their respective neighborhood.


Spatialized indicators are often subject to spatial dependencies (or interactions), which are even stronger than spatial locations are closer. Autocorrelation measures (Moran, Geary, Lisa indexes) allows estimating the spatial dependence between the values of a same indicator at several locations of a given study area.

The figure displayed below consists in evaluating this spatial autocorrelation by a plot crossing the spatial deviation (Y axis) and global deviation (X axis) values.

This plot provides interesting inputs for answering to basic questions, such as: Are territorial units in favorable situation in a global context also in the same situation at local level? Is it possible to detect specific situations as regard to the global trend, meaning the existence of "local bastions" in favorable or lagging situations?

opar <- par(cex.lab = 1, cex.axis = 0.75, mar = c(4, 4, 1.2, 2))

# Drop geometries
df <- st_set_geometry(com, NULL)

# Spatial autocorrelation
lm <- summary.lm(lm(sdevrel ~ gdevrel, df))

# Equation 
eq <- paste("Spatial Deviation =", 
            round(lm$coefficients["gdevrel","Estimate"], digits = 3),
            "* (Global Deviation) +", 
            round(lm$coefficients["(Intercept)","Estimate"], digits = 3))
rsq <-paste("R-Squared =", 
            round(summary(lm(sdevrel ~ gdevrel, com ))$r.squared, digits = 2))

# Color management
df <- merge(df, colEpt, by = "LIBEPT")

# Plot spatial autocorrelation
plot(df$gdevrel, df$sdevrel,
     ylab = "Local deviation",
     ylim = c(50,260),
     xlab = "Global deviation",
     xlim = c(50,260),
     pch = 20,
     col = as.vector(df$col),
     asp = 1)
abline((lm(df$sdevrel ~ df$gdevrel)), col = "red", lwd =1)

# Specify linear regression formula and R-Squared of the spatial autocorrelation
text(110,60, pos = 4, cex = 0.7, labels = eq)
text(110,55, pos = 4, cex = 0.7, labels = rsq)

abline (h = seq(40,290,10), col = "gray70", lwd = 0.25, lty = 3)
abline (h = seq(50,250,50), col = "gray0", lwd = 1, lty = 1)
abline (v = seq(40,290,10), col = "gray70", lwd = 0.25, lty = 3)
abline (v = seq(50,250,50), col = "gray0", lwd = 1, lty = 1)

# Legend for territorial level
legend("topleft", legend = rev(as.vector(colEpt$LIBEPT)), pch = 15,
       col = rev(as.vector(colEpt$cols)), cex = 0.6, pt.cex = 1.5)
par(opar)

The output of the linear model of spatial autocorrelation reveals that the hypothesis of independence is rejected at a probability below than 0.0001. It means that, "everything is related to everything else, but near things are related than distant things" (Tobler, 1970). However, the R-squared of the relation (0.42) suggests that this statistical relation includes outliers very far from the linear regression.

This chart can be modified for analyzing outliers specifically. The code below computes the statistical residuals of the spatial autocorrelation calculated above. Then the residuals are standardized. Finally, a plot is built to display and label significant residuals.

opar <- par(cex.lab = 1, cex.axis = 0.75, mar = c(4, 4, 2, 2))

# Standardized residual calculation
lm <- lm(sdevrel ~ gdevrel, df)
df$res <- rstandard(lm)

#risk alpha (0.1 usually)
alpha <- 0.055

# Calculation of the threshold using T-Student at (n-p-1) degrees of freedom
thr <- qt(1 - alpha / 2, nrow(com) - 1)

# Plot residuals
plot(df$sdevrel, df$res,
     xlab = "Local deviation", cex.lab = 0.8,
     ylim = c(-3.5, 3.5),
     xlim = c(40, 200),
     ylab = "Standardized residuals of spatial autocorrelation", 
     cex.lab = 0.8,
     cex.axis = 0.8,
     pch = 20,
     col = as.vector(df$cols))

# Adding thresholds
abline(h = - thr, col = "red")
abline(h = + thr, col = "red")
abline(h = 0, col = "red")

# Detecting exceptional values and labeling them on the plot
ab <- df[df$res < -thr | df$res > thr,]

# Plot residual labels
text(x = ab[,"sdevrel"], y = ab[,"res"], ab[,"LIBCOM"], cex = 0.5, pos = 4)

abline (v = seq(50, 200, 10), col = "gray70", lwd = 0.25, lty = 3)
abline (v = seq(50, 200, 50), col = "gray0", lwd = 1, lty = 1)

# Plot the legend (territorial zoning)
legend("topleft", legend = rev(as.vector(colEpt$LIBEPT)), pch = 15,
       col = rev(as.vector(colEpt$cols)), cex = 0.6, pt.cex = 1.5)
par(opar)

Further analysis may be considered in the domain, using for instance Moran indexes et LISA indexes or other distance criteria. This, it is important to remind that the choice of the appropriate threshold or criteria must make sense from a thematic point of view (does a neighborhood of 5 km mean something? Is a distance of 20 minutes by car more appropriated?)


3. Redistributions

Redistributions highlight another way to analyze territorial disparities. The analysis is no more focused on inequalities themselves, but rather than to the statistical process (amount of numerator or denominator) required to reach the equilibrium in each territorial context.

The meaning of this mathematical calculation is strong in policy term, since it point out the municipalities which may have to contribute / receive the highest to ensure a perfect equidistribution in several territorial contexts. It gives also the amount of the reallocation process.

The calculation is made for the general and the territorial contexts, using the argument type = abs. Two maps are created, displaying the amount of income that should be transfered from the wealthiest to the poorest municipalities.

# general absolute deviation 
com$gdevabs <- gdev(x = com, 
                    var1 = "INC", 
                    var2 = "TH", 
                    type = "abs")

# Territorial absolute deviation calculation
com$tdevabs <- tdev(x = com, 
                    var1 = "INC", 
                    var2 = "TH", 
                    type = "abs",
                    key = "LIBEPT")

# Transform the values in million Euros
com$gdevabsmil <- com$gdevabs / 1000000
com$tdevabsmil <- com$tdevabs / 1000000

# Deviation orientation
com$gdevsign <- ifelse(com$gdevabsmil> 0, "Income surplus", "Income deficit")
com$tdevsign <- ifelse(com$tdevabsmil > 0, "Income surplus", "Income deficit")

# Deviation maps 
opar <- par(mfrow = c(1,2))

# General deviation
# Plot territories
mf_map(com, col = "peachpuff", border = "white", lwd = 0.25)
mf_map(ept, col = NA, border = "#1A1A19", lwd = 1, add = TRUE)

mf_map(x = com, var = c("gdevabsmil", "gdevsign"), type = "prop_typo",
       leg_title = c("Absolute Deviation\n(Income redistribution, euros)",
                    "Redistribution direction"),
       leg_pos = c("topleft", NA), leg_val_rnd = -2,
       val_order = c("Income deficit", "Income surplus"),
       pal =  c("#ff0000","#0000ff"), border = "grey70", 
       add = TRUE)

# Labels to ease comment location
mf_label(x = com[com$LIBCOM %in% c("Paris 7e Arrondissement",
                                   "Neuilly-sur-Seine", "Aubervilliers") ,],
         var = "LIBCOM", cex = 0.6, font = 2,
         halo = TRUE, overlap = FALSE)

# Layout map 1
mf_layout(title = "General deviation (Metrople du Grand Paris)",
          credits = paste0("Sources : GEOFLA® 2015 v2.1, Apur, impots.gouv.fr",
                           "\nRonan Ysebaert, RIATE, 2021"), 
          arrow = FALSE)

# Territorial deviation
mf_map(com, col = "peachpuff", border = "white", lwd = 0.25)
mf_map(ept, col = NA, border = "#1A1A19", lwd = 1, add = TRUE)

mf_map(x = com, var = c("tdevabsmil", "tdevsign"), type = "prop_typo",
       leg_title = c("Absolute Deviation\n(Income redistribution, euros)",
                    "Redistribution direction"),
       leg_pos = c(NA, "topleft"), leg_val_rnd = -2,
       val_order = c("Income deficit", "Income surplus"),
       val_max = max(abs(com$gdevabsmil)),
       pal =  c("#ff0000","#0000ff"), border = "grey70", 
       add = TRUE)

# Labels to ease comment location
mf_label(x = com[com$LIBCOM %in% c("Marnes-la-Coquette",
                                   "Nanterre", "Clichy-sous-Bois") ,],
         var = "LIBCOM", cex = 0.6, font = 2,
         halo = TRUE, overlap = FALSE)

# Layout map 2
mf_layout(title = "Territorial deviation (EPT of belonging)",
          credits = paste0("Sources : GEOFLA® 2015 v2.1, Apur, impots.gouv.fr",
                           "\nRonan Ysebaert, RIATE, 2021"), 
          arrow = FALSE)
par(opar)

In the context of the MGP, the 7th Arrondissement of Paris is the municipality which should contribute the most, all things being equal to its income tax level (1,987 billion Euros of income transfer, 65 % of the total amount of tax income in this municipality). Neuilly-sur-Seine (third position in absolute terms) should transfer 1,9 billion Euros to the lagging municipalities of the MGP area. It corresponds to 62,87% of the total amount of tax income declared in this municipality.

In the other side of the redistribution, La Courneuve should receive 402 million Euros from the wealthiest municipalities (127 % of the current tax income declared). Aubervilliers should receive 793 million Euros, which represents 124 % of the current tax income in this municipality.

The chunks below compute the top 10 theoretical contributors and receivers municipalities (total amount of income and share of current available income) to reach a perfect equilibrium.

# general deviation - Top 10 of the potential contributors in regard 
# Drop geometries
df <- st_set_geometry(com, NULL)
row.names(df) <- df$LIBCOM

# to their total amount of income
df$gdevabsPerc <- df$gdevabs / df$INC * 100
df <- df[order(df$gdevabsPerc, decreasing = TRUE), ]
df[1:10, c("gdevabsmil","gdevabsPerc")]

# general deviation - Top 10 of the potential receivers in regard to 
# their total amount of income
df <- df[order(df$gdevabsPerc, decreasing = FALSE), ]
df[1:10, c("gdevabsmil", "gdevabsPerc")]

Looking at the EPT context, the 7th Arrondissement of Paris is still the municipality which should contribute the most to the poorest municipalities of Paris as regards to the amount of income available in this municipality (1,779 billion Euros of income transfer, 58 % of the amount of income in this municipality). Marnes-La-Coquette (second position) should transfer 38 million Euros to the poorest municipalities of its EPT of belonging (La Défense). Despite the low income mass, it corresponds to 52 % of the total amount of income available in this municipality.

From the other side of the redistribution, Nanterre, Clichy-sous-Bois and the 19th Arrondissement of Paris should receive respectively 1088, 143 and 1926 million euros from the wealthiest municipalities of their EPT of belonging (respectively 88 %, 68 % and 65 % of the total amount of earned income of their households). The highest redistribution for this study area stands for the 20th Arrondissement of Paris (1,926 billion Euros, 59 % of its total amount of income).

# general deviation - Top 10 of the potential contributors in regard 
# Drop geometries
df <- st_set_geometry(com, NULL)
row.names(df) <- df$LIBCOM

# Territorial deviation - Top 10 of the potential contributors
# as regards to their total amount of income
df$tdevabsPerc <- df$tdevabs / df$INC * 100
df <- df[order(df$tdevabsPerc, decreasing = TRUE), ]
df[1:10, c("tdevabsmil", "tdevabsPerc")]

# Territorial deviation - Top 10 of the potential receivers 
# as regards to their total amount of income
df <- df[order(df$tdevabsPerc, decreasing = FALSE), ]
df[1:10, c("tdevabsmil", "tdevabsPerc")]


4. Synthesis

This section shows the synthesis functions of the MTA package: bidev, plot_bidev and map_bidev are useful to synthetize, plot and map the situation for 2 selected relative deviations. The mst, plot_mst and map_mst are suitable for 3 deviations.

4.1 Two-relative deviations synthesis

The bidev function allows firstly to position each territorial unit as regards to 2 deviations (dev1 and dev2) already calculated; and secondly to evaluate the statistical distance to the average (index 100). This function returns a vector which is the combination of two modalities :

Are combined to modalities A, B, C and D the distance to the average. By default, it corresponds to 25 %, 50 % and 100 % under-above the average. Adapted to normalized index where 100 is average, it creates 4 classes :

At this end, it produces a 13 classes-typology, which is the combination of the two modalities described above (classes A1, B3, C2, etc.). The user is free to modify the breaks proposed by default by the function (breaks = c(25, 50, 100))

To understand the resulting typology, the plot_mst is especially adapted. The X-Y scale is defined in logarithm. Index 200 corresponds to territorial units defined to twice the average ; and index 50 twice below the average.

opar <- par(mar = c(0, 0, 0, 0))

# Prerequisite  - Compute 2 deviations
com$gdev <- gdev(x = com, var1 = "INC", var2 = "TH")
com$tdev <- tdev(x = com, var1 = "INC", var2 = "TH", key = "EPT")

# EX1 standard breaks with four labels
plot_bidev(x = com, 
           dev1 = "gdev", 
           dev2 = "tdev",
           dev1.lab = "General deviation (MGP Area)",
           dev2.lab = "Territorial deviation (EPT of belonging)",
           lib.var = "LIBCOM",
           lib.val =  c("Marolles-en-Brie", "Suresnes", 
                        "Clichy-sous-Bois", "Les Lilas"))
par(opar)

This plot highlights specific areas of interest. In the red quarter, territorial units above the average for the general and territorial deviations, such as Marolles-en-Brie(class A3). Reversely, municipalities appearing in the blue quarter are below the average for the 2 deviations, such as Clichy-sous-Bois(class C3).

It shows also territorial units in contradictory situations, such as Les Lilas(class D2, green quarter), which is in lagging situation for the general deviation and in favorable situation for the territorial context ((EPT of belonging). Reversely, Suresnes is in favorable situation in the context of the MGP and in lagging situation in its EPT of belonging.

The map_bidev function delivers a list of 3 objects useful to map the results: a sf object (geom) including the result of the bidev function (new column). This object is ordered following the 13 classes. It delivers also a color vector for mapping purpose. It includes only resulting categories: if the dataset does not include "D3" class, the resulting cols object will not include dark green color.

For mapping the results, it is recommended to share the plot in two columns : one side for the map, and one side displaying the bidev plot to understand correctly the colors displayed on the map.

opar <- par(mfrow = c(1,2))

bidev <- map_bidev(x = com, dev1 = "gdev", dev2 = "tdev",
                   breaks = c(50, 100, 200))

# Unlist resulting function
com <- bidev$geom
cols <- bidev$cols

# Cartography
mf_map(x = com, var = "bidev", type = "typo", 
       pal = cols, lwd = 0.2, border = "white", leg_pos = NA,
       val_order = unique(com$bidev))

mf_map(ept, col = NA, border = "#1A1A19", lwd = 1, add = TRUE)

# Label territories in the C3 category
mf_label(com[com$bidev == "C3",],
         var = "LIBCOM", halo = TRUE)

mf_title( "2-Deviations synthesis:", tab=FALSE)
mf_layout(title = "",
          credits = paste0("Sources : GEOFLA® 2015 v2.1, Apur, impots.gouv.fr",
                           "\nRonan Ysebaert, RIATE, 2021"), 
          arrow = FALSE)
mf_title("general and territorial contexts", tab = FALSE, inner = T)

#Associated plot
plot_bidev(x = com,  dev1 = "gdev",  dev2 = "tdev", 
           dev1.lab = "General deviation (MGP Area)",
           dev2.lab = "Territorial deviation (EPT of belonging)",
           lib.var = "LIBCOM", lib.val = "Clichy-sous-Bois", cex.lab = 0.8)
par(opar)

This map highlights areas in favorable situation for the MGP area and EPT contexts (red), in lagging situation for these two contexts (blue) or in contradictory situations (yellow and green). Darker is the color, stronger are the distance to the average.

4.2 Three-relative deviations synthesis

The three relative deviations (general, territorial, spatial) can be summarized in a synthetic typology using the mst function. It allows, according to a pre-defined threshold (100 being the average), to higlight the territorial units which are under or above this threshold, and for which of the 3 deviations proposed by the MTA package.

Assuming (for instance) that the treshold is equal to 100 (value corresponding to the average for each deviations) and the superior argument is false (consider the value below the average), the mst() function returns a 8-classes typology (mst column) which can be interpreted as follows:

As demonstrated below, this typology is especially useful to highlight territories in lagging or favorable situations, but also territories in contradictory situations.

4.2.1 Wealthiest municipalities

We focus firtstly the analysis on municipalities above 125 % for the three deviations (gdev = Métropole du Grand Paris ; tdev = EPT average and sdev = contiguous municipalities, threshold = 125, superior = TRUE). The mst() function returns their position according to the three deviations. The subset (mst == 7) returns all the municipalities in this situation.

# Prerequisite  - Compute the 3 deviations
com$gdev <- gdev(x = com, var1 = "INC", var2 = "TH")
com$tdev <- tdev(x = com, var1 = "INC", var2 = "TH", key = "EPT")
com$sdev <- sdev(x = com, var1 = "INC", var2 = "TH", order = 1)

# Multiscalar typology - wealthiest territorial units
# Row names = municipality labels
row.names(com) <- com$LIBCOM

# Compute mst
com$mstW <- mst(x = com, gdevrel = "gdev", tdevrel = "tdev", sdevrel = "sdev", 
                threshold = 125, superior = TRUE)

subset(com, mstW == 7, select = c(ratio, gdev, tdev, sdev, mstW), drop = T)

The results are afterwards mapped. As map_bidev function do, the map_mst function takes in entry the 3 pre-calculated deviations and mst parameters and returns a list including a sf object with the result of the mst calculation, and two vectors proposing colors and legend for mapping results.

# Compute mapmst
mst <- map_mst(x = com, gdevrel = "gdev", tdevrel = "tdev", sdevrel = "sdev", 
               threshold = 125, superior = TRUE)

# Unlist resulting function
com <- mst$geom
cols <- mst$cols
leg_val <- mst$leg_val

# Cartography
mf_map(x = com, var = "mst", type = "typo", 
       border = "white", lwd = 0.2,
       pal = cols, val_order = unique(com$mst), leg_pos = NA)
mf_map(ept, col = NA, border = "black", lwd = 1, add = TRUE)
mf_legend(type = "typo", pos = "topleft", val = leg_val, pal = cols, 
          title = paste0("Situation on General (G)\n",
                         "Terrorial (T) and \n",
                         "Spatial (S) contexts"))
mf_layout(title = "3-Deviations synthesis: Territorial units above index 125",
          credits = paste0("Sources : GEOFLA® 2015 v2.1, Apur, impots.gouv.fr",
                           "\nRonan Ysebaert, RIATE, 2021"), 
          arrow = FALSE)

# Add labels for mst = 7
mf_label(x = com[com$mst == 7,], var = "LIBCOM",  halo = TRUE, overlap = FALSE, 
         cex = 0.7)

This map highlights the municipalities 25 % above the average for the three deviations (colored in red). They are mainly located in the West part of Paris and of the MGP. Some isolated municipalities appear also locally in a favorable situation in the East Part of the MGP.

More precisely, 11 municipalities are characterized by an index of average income per household above 125% on the three contexts. It corresponds to 4 Arrondissements in Paris (6e, 7e, 8e and 16e) and the municipalities of Marnes-la-Coquette, Neuilly-sur-Seine, Sceaux, Vaucresson, Marolles-en-Brie, Saint-Mandé and Santeny.

13 municipalities can be considered as "globally" advantaged (above 125% only in a global and/or in a global and in territorial contexts, colored in light orange and in pink). It corresponds to the central Arrondissements in Paris (1st, 2nd, 4th, 9th) and other municipalities in favorable situation close to wealthier municipalities (Boulogne-Billancourt, Saint-Cloud, etc).

8 municipalities can be considered as "locally" advantaged (index above 125% only in a spatial context and/or in a spatial and a territorial context, colored in green and yellow). These local poles of wealth are mainly located in the periphery of the MGP area (Le Raincy, Rungis, Coubron, etc.), closed to poorer municipalities.

4.2.2 Municipalities in lagging situation

It is also possible to analyze the reverse situation: the municipalities in lagging situation for this indicator. Thus, the map_mst() function takes '80' for the threshold argument and 'FALSE' for the superior argument. This typology will specify the municipalities situated below 25 % of the average for the global and/or the territorial and/or the spatial contexts.

# Compute mapmst
mst <- map_mst(x = com, gdevrel = "gdev", tdevrel = "tdev", sdevrel = "sdev", 
              threshold = 80, superior = FALSE)

# Unlist resulting function
com <- mst$geom
cols <- mst$cols
leg_val <- mst$leg_val

# Cartography
mf_map(x = com, var = "mst", type = "typo", 
       border = "white", lwd = 0.2,
       pal = cols, val_order = unique(com$mst), leg_pos = NA)
mf_map(ept, col = NA, border = "black", lwd = 1, add = TRUE)
mf_legend(type = "typo", pos = "topleft", val = leg_val, pal = cols, 
          title = paste0("Situation on General (G)\n",
                         "Terrorial (T) and\n",
                         "Spatial (S) contexts"))
mf_layout(title = "3-Deviations synthesis: Territorial units under index 80",
          credits = paste0("Sources : GEOFLA® 2015 v2.1, Apur, impots.gouv.fr",
                           "\nRonan Ysebaert, RIATE, 2021"), 
          arrow = FALSE)

# Add labels for mst = 7
mf_label(x = com[com$mst == 7,], var = "LIBCOM",  halo = TRUE, overlap = FALSE, 
         cex = 0.7)

This map shows the municipalities in difficulty regarding the income distribution in the MGP. Interesting is to note that all the territories of the Plaine Commune, Territoire des Aéroports EPT are below the threshold of 80 % for at least the global context. Most of the municipalities of EPT of Est Ensemble, Grand-Paris-Est, Val de Bièvres are in the same situation.

The municipalities in lagging situation for the three contexts are located in several areas of the MGP (North-West, East). All are not located not in the immediate periphery of Paris (none of these municipalities are contiguous to Paris). It concerns precisely 6 municipalities: Bagneux, Genevilliers, Nanterre, Clichy-sous-Bois, Bonneuil-sur-Marne and Champigny-sur-Marne. Clichy-sous-Bois appears especially in a dramatic situation (indexes 47, 60 and 61 for the global, territorial and spatial deviations).

Most of the municipalities of Seine-Saint-Denis and Val-de-Marne are characterized by a significant lagging situation globally but not locally (classes 1 and 3, in light orange and pink).

# Multiscalar typology - Lagging territorial units
# Row names = municipality labels
row.names(com) <- com$LIBCOM

# Compute mst
com$mstP <- mst(x = com, gdevrel = "gdev", tdevrel = "tdev", sdevrel = "sdev", 
                threshold = 80, superior = FALSE)

# municipalities in lagging situation for the three contexts
subset(com, mstP == 7, select = c(ratio, gdev, tdev, sdev, mstP), drop = T)

4.2.3 Contradictory situations

As displayed on the map above, some territories appear also in a contradictory situation depending on the deviation used.

Municipalities characterized by class 3 in the mst synthesis are lagging as regards to the global and the territorial contexts, but not the spatial one. It is especially the arrondissements located in the North-West part of Paris, contiguous to poorest municipalities.

# municipalities in lagging situation in the global and territorial contexts
subset(com, mstP == 3, 
       select = c(ratio, gdev, tdev, sdev, mstP), drop = T)
# municipalities in favorable situation in a spatial context or in a spatial and a territorial context 
subset(com, mstP == 4 | mstP == 6, 
       select = c(ratio, gdev, tdev, sdev, mstP), drop = T)

4 municipalities are characterized by low indexes in a local context, but not in a global one. It corresponds to the municipalities located in the immediate neighborhood of the wealthiest municipalities of the MGP (classes 4 and 6, in green and yellow): Paris 15e Arrondissement, Suresnes, Puteaux and in a lesser extent Malakoff.

4.2.4 Comparing specific territorial units

The plot_mst allows to compare selected territorial units on a barplot. This function takes in entry 3 pre-calculated deviations and returns a barplot which shows the position on the three deviations for one (or more) territorial unit(s).

opar <- par(mar = c(4, 6, 4, 4))
# Synthesis barplot
plot_mst(x = com, 
         gdevrel = "gdev", tdevrel = "tdev", sdevrel = "sdev", 
         lib.var = "LIBCOM", 
         lib.val = c("Neuilly-sur-Seine", "Clichy-sous-Bois", 
                     "Suresnes", "Les Lilas"))
par(opar)

The barplot above displays the situation for 4 municipalities of the study area ( Les Lilas, Neuilly-sur-Seine, Suresnes, Clichy-sous-Bois) is a good way to propose a visualization for comparing specific territorial units.

4.3 Absolute deviations synthesis

The mas() function takes in entry all requested parameters to compute the three deviations, as specified above. It returns a dataframe summarizing the values of absolute deviations, e.g. how much should be redistributed from the poorest to the wealthiest territorial units to ensure a perfect equilibrium of the ratio for the three contexts. Results are expressed both in absolute values (mass of numerator, amount of tax reference in Euros in this case) and as a share of the numerator (x % of the numerator that should be redistributed).

# Local redistribution (not yet calculated)
com$sdevabs <- sdev(x = com, xid = "DEPCOM", var1 = "INC", var2 = "TH",
                    order = 1, type = "abs")
com$sdevabsmil <- com$sdevabs / 1000000

# Compute the synthesis DataFrame (absolute deviations)
mas(x = com, 
    gdevabs = "gdevabsmil", 
    tdevabs = "tdevabsmil",
    sdevabs = "sdevabsmil",
    num = "INCM") 

For the MGP area, it is 22 billion Euros that should be redistributed from the municipalities in favorable situation to the municipalities in lagging situation. It corresponds to 16.3 % of the total mass of income declared to the taxes. If a policy option consists in ensuring a equilibrium at an intermediate territorial level, such as the Établissements Publics Territoriaux, it is 15 billion Euros that should be redistributed (11.6 % of the income mass). If a solution chosen consists in limiting territorial discontinuities in a local context (avoid local poles of wealth or of poverty), it is 10 billion Euros that should be redistributed (7.9 % of the income mass). Ensuring an equilibrium of income in these three territorial contexts are obviously not credible policy options, but it gives some references to monitor the magnitude of territorial inequalities existing in a given study area.


5. Conclusion and perspectives

The functionalities provided by the MTA package are especially useful to enhance territorial inequalities in several perspectives. This example, applied to the Métropole du Grand Paris provides some interesting insights to understand challenges raised by income inequalities in this governance structure at the scale of the Parisian metropolitan area. Inputs provided by this vignette are quite simple and aims at using all the functionalities of the MTA package, by suggesting some adapted graphical representations. We consider also that it provides useful inputs for territorial monitoring and observation - a necessary step before policy decision making (which solutions for reducing income inequalities in a given area?)

It is also important to remind that the aim of this vignette is not to provide a global picture of territorial inequalities existing in the MGP. To have this ambition, further analysis should be led for interrogating other dimensions of the inequalities, such as:

That being said, MTA analysis provide a useful methodological base to explore several dimensions of territorial inequalities measures and applications within a policy context:



Try the MTA package in your browser

Any scripts or data that you put into this service are public.

MTA documentation built on Nov. 2, 2023, 5:06 p.m.