Multidimensional deprivation in heterogeneous rural areas: Spain after the economic crisis

ABSTRACT

We analyse the changes in multidimensional deprivation in very heterogeneous rural areas in Spain during the 2008 economic crisis using multigroup latent class models. Counterfactual distributions are implemented to identify the factors behind the change in deprivation in the different areas. We find that the economic crisis negatively affected direct indicators of the living standards in rural areas. A wide range of differences appears when specific rural areas are studied going beyond the usual dilemma between rural and urban areas. The results also belie the common stereotype that the greatest incidence of monetary poverty in rural areas is offset by better living conditions.

KEYWORDS:

• rural areas
• multidimensional deprivation
• latent class model
• European Union Statistics on Income and Living Conditions (EU-SILC)

JEL:

• I32
• R13

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Notes

1. The year 2005 was chosen because it is the only year before the crisis for which the National Institute of Statistics could make the disaggregation of rural areas requested by the authors. Income data collection in the survey completely changed in 2013 and the 2012 wave was chosen for the sake of comparability.

2. A two-stage design is used with first-stage unit stratification. For each region, the first-stage units are grouped into strata in agreement with the size of the municipality to which the section belongs. The sections are selected within each stratum with a probability proportional to their size.

3. Pereira et al. (2004) use the territorial unit called ‘comarca’, which is a division of territory that usually coincides with a natural area that shares physical (orography, hydrography, climate, vegetation and soil), human (demography, economic uses, rural housing and urban planning) and historical characteristics, all of which determine its geographical landscape. This unit is much smaller than NUTS-3 units. Therefore, there is greater homogeneity regarding productive activity, geographical, economic and social characteristics.

4. Only one area can be assigned to each municipality, so they are mutually exclusive categories.

5. No other scientific study classifies the rural world in Spain using such small territorial units and so many variables related to geography, demography and economy.

6. The number of observations in the survey for each area is as follows: urban areas (20,920), intermediate areas (5726), arable crops and permanent pastures smallholdings (2313), arable crops and permanent pastures large holdings (2252), and mountain areas (1639). Since the survey is representative at the regional level and in more than half the regions the sample is smaller than in the rural area with the lowest number of observations, it can be expected that the defined areas are sufficiently representative.

7. In addition, tests of equality of means between the two years in these binary variables were conducted for each area considering the year as a group variable. Age, gender, size and some types of household can be considered homogeneous according to these tests, finding in all groups – as in the country as a whole – higher educational levels in all areas in 2012.

8. For a formalization of the model, see Pérez-Mayo (2005).

9. This structural model will be estimate twice, one for each year. The spatial areas – and not the years – define the groups.

10. Given the randomly distribution of missing data, the gross indicators for these cases are recoded as non-deprived.

11. Housing cost overburden exists if costs > 40% of disposable household income without any possible housing allowances. Overcrowding happens if there are not: (1) one room for the household; (2) one room per couple; (3) one room for each single person 18 years of age or older; (4) one room per pair of single people of the same gender between 12 and 17 years of age; (5) one room for each single person of different gender between 12 and 17 years of age; and (6) one room per pair of children under 12 years of age.

12. Appendix B in the supplemental data online provides a detailed explanation of goodness-of-fit testing for latent class models.

13. Since households in both waves are different, the model was estimated again in order to find the best one. Therefore, response probabilities are allowed to be different.

14. Due to space constraints, it was not possible to include a detailed comparison of the results using national and area reference frameworks. Using national instead of area-specific references produces a higher incidence of deprivation. The results with both options are available from the authors upon request.

15. These estimates are available from the authors upon request.

Additional information

Funding

Luis Ayala gratefully acknowledges funding from the Comunidad de Madrid [grant number H2019/HUM-5793] and from the Spanish Ministerio de Economía y Competitividad [grant number ECO2016-76506-C4-3-R]. Antonio Jurado and Jesús Pérez-Mayo gratefully acknowledge funding from the Consejería de Educación y Empleo, Junta de Extremadura and the ERDF funds [grant number GR15023].