Fertility along the Paris Metro A visualization and multilevel - - PDF document

1

Fertility along the Paris Metro A visualization and multilevel analysis of local fertility differences in Paris

John Tomkinson (University of Strasbourg / INED)

Introduction

The city of Paris distinguishes itself from the rest of mainland France due to its low level of fertility having a total fertility rate of 1.6 children per woman compared to 2.0 in mainland France as a whole. The fertility calendar is also later, with mean age at birth being nearly 3 years older than mainland France (32.9 years

• ld vs 30.0) (Dubujet, 2013).

These statistics hide the vast variations in fertility at local level within the city of Paris. This poster has two objectives: (i) To highlight the variation in levels of fertility within the city of Paris by presenting in an

• riginal and innovative way the results of local level analysis. To raise awareness amongst the

general public, this poster takes inspiration from Cheshire’s “Lives on the Line” (2012) and calculates total fertility rates and mean age at childbearing for each of Paris’ 245 metro

• stations. The aim is to highlight spatial variability of fertility in a context that the general

public understands – for Parisians, metro stations serve not only as a means of transport but as local landmarks, and even for non-Parisians, serve as a far more pertinent geographical marker than traditional administrative zones. The poster aims to place statistics in a geographical context with real meaning. (ii) To investigate the individual and contextual factors behind the spatial variation of fertility in Paris by conducting a multilevel analysis using individual level French census data – regrouping information on more than half a million women of reproductive age – and a specific application of the Own Children method, improved for application on the French census.

Methods and Data

Data The source of data used for the analyses is the individual level database of the French census of 2011, containing sociodemographic data upon each individual enumerated in the city of Paris. Using a specially adapted version of the Own Children method (Tomkinson and Breton, 2017), we can link mothers to their children within a household. Once successfully linked to their children, it is easy to calculate the age of the woman in order to calculate the mean age at childbearing and each linked mother-child is counted in the numerator (equivalent to a birth) in order to calculate total fertility rate. The sociodemographic data relating to the mother, present in the census yet absent in birth registration data, allows us to examine the individual and contextual factors associated with the probability to have given birth to a child in the year prior to the census.

2 Crucially, the individual level census data also contains a fine level geographical allowing us to localize the births. The variable IRIS corresponds to a census tract (existing in all French cities of at least 10,000 habitants) of which there are 992 in the city of Paris. Methodology Using a method similar to that of Cheshire (2012) to calculate life expectancy, Paris’ metro stations are geolocalised and superimposed upon the map of Paris’ IRIS (Figure 1). A circle with a radius of 500m (distance equivalent to 5 minutes’ walk) is traced around each station and each IRIS covered by this zone is used in the calculation of the fertility rate and mean age at birth of the station. Thus, to calculate indicators for station 1 – Gare du Nord – we use data from the IRIS A, B, C, D, E, F, G, H, I, J and K, and for station 2 – Gare de l’Est – from the IRIS E, F, I, J, K, L, M, N, O, P and Q. In general we thus calculate total fertility rate at station s (TFRs) as: 𝑈𝐺𝑆𝑡 = 5 ∑ 𝑔

𝑦 45−49 15−19

Where the fertility rate for 5 year age groups (fx) is: 𝑔

𝑦 =

∑ 𝑐𝑦

𝑜 𝑗

∑ 𝑞𝑦

𝑜 𝑗

with bx = births1 in the age group x, px = number of women in the age group x and n the number

• f IRIS in a radius of 500m of the station s.

Similarly, we calculate the mean age at birth at station s (Ms) as: 𝑁𝑡 = ∑ ∑ 𝑐𝑙

𝑛 𝑙 𝑜 𝑗

∑ 𝑐

𝑜 𝑗

with bk = births at age k, k = minimum age at birth and m = maximum age at birth.

1 As the Own children method underestimates the number of births, we correct the number of births using a

‘correctional weight’ which differs according to the arrondissement in which the IRIS is located and 5 year age group. This corresponds to the number of births in an arrondissement in a 5 year age group according to vital statistics, divided by the same number according to estimates from census data.

3

Results

 Cartography TFR varies widely from a low of 0.94 at the station Maubert-Mutualité to a high of 2.38 at Porte de

• Pantin. The MACB ranges from 30.6 (Bercy) to 34.7 years old (École Militaire). Both TFR and MACB

change along metro lines traversing the city and can vary widely between two consecutive stations Figure 2 shows an extract of the mapped total fertility rates and mean age at birth at each metro

• station. This extract shows how TFR increases along the line 11 (brown line) as it heads from central Paris

to the North Eastern suburbs (cf. Figure 3). At Châtelet the TFR is 1.31 children per woman compared to 1.96 at Porte des Lilas.

• Fig. 2 – Extract of poster showing total fertility rates (TFRs) and mean age at childbirth (Ms) mapped to

Metro stations

Sources: Insee RP 2013 & Etat civil, IGN, OpenStreetMap.

• Fig. 3 – Variation of total fertility rates (TFRs) along the Line 11

Sources: Insee RP 2013 & Etat civil, IGN, OpenStreetMap.

4  Multilevel analysis A multilevel analysis using individual level data (age, education level, economic activity, living in scoail housing, immigration status) and aggregate local level data (unemployment rate, proportion of single parent families, proportion professional workers, proportion manual workers, proportion of

• vercrowded housing, proportion of non-home owners) shows that individual characteristics explain 95%
• f the variation in TFR.

The strongest associated factors are age – women aged under 20 and 45 and over are less likely to have had a child in the last year compared to women aged 30-34 (β = -1.8 and -3.7 respectively), being in a couple (β = 2.3). Women economically inactive are also more likely to have had given birth in the last year compared to employed women (β = 0.6).

Bibliography

Cheshire, J. 2012. “Featured Image: Lives on the Line.”, Environment and Planning A, 44(7), p. 1525-1528. Dubujet, F. 2013. “Ile-de-France : une fécondité toujours élevée, des naissances de plus en plus tardives”, Ile-de-France – Faits et chiffres, n° 299, 3 p. Tomkinson, J. & Breton D. 2017. ““Comment mieux identifier les mères adolescentes dans le recensement français ? Améliorations de la méthode du “décompte d’enfants au foyer””, Cahiers québécois de démographie (in press).