Abstract
This chapter presents the last set of analyses carried out in the framework of this research. In the first part of the chapter, I provide the specific hypotheses, formulated based on the objectives of the research. The second part presents some descriptive bivariate analyses, similar to those presented in Chap. 3, where the information on the features of the domestic sector is aggregated at the level of the typologies, instead of the countries. The third part of the chapter presents the final analysis, the comparison between the different models that I have estimated and the interpretation of the results.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
From a statistical point of view, combining data on individuals with information aggregated at a higher level (the level of the group) is a situation that needs to be addressed with specific tools. When using statistical methods, the effect that the group has on the individual level should be taken into account, to rule out the possibility that the observed between-individual variability is in fact a between-group variability. Multilevel analysis is considered the most adequate statistical tool to analyse hierarchical data, as it is specifically designed to acknowledge the presence of a hierarchical structure in the data. However, as most multivariate techniques, it cannot be applied to all situations, as it requires a series of assumptions to be met. Specifically, because multilevel analysis requires the sample size, at both the individual and the group levels, be sufficiently large, it is not suited to comparisons between European countries, the number of groups (i.e. countries) being limited to a maximum of 28 (for a discussion on the ‘rules of thumb’ for the use of multilevel analysis, see Hox [1998], Twisk [2006], Maas and Hox [2005], Bryan and Jenkins [2015]). Therefore, multilevel analysis could not be used in this research instance, since the number of clusters and/or countries is too small.
- 2.
The discussion on the definition of the migrant population using the EU-LFS data and its limitations is provided in Chap. 1.
- 3.
The justification and the limitations of using these ISCO-08 codes for the definition of the domestic sector are thoroughly discussed at the beginning of Chap. 3.
- 4.
The variable that allows to identify migrants (‘country of birth’) is only partially available for Germany, due to national restrictions for researchers. Specifically, instead of providing information on the origin of the individual (country areas), the available variable only distinguishes between the category ‘own country’ and ‘no answer’. After cross-checking the missing information (‘no answer’) for this variable in all the other countries in the database, I decided to use the category ‘no answer’ as representative of the migrant population. Indeed, in all countries, the missing information (‘no answer’) represents a very small percentage of the responses (for instance, only 36 ‘no answer’ for Belgium, 26 for Luxembourg, 178 for Sweden, 18 for Portugal, while the majority of countries have no missing responses). Therefore, even if this category might include some “real” missing information, I assume that it nevertheless represents a good approximation of the number of migrants for Germany in this database.
- 5.
The original variable is coded into the eight levels of the ISCED 11 classification. The new variable is recoded into: Category 1 (low education), which groups together old categories ‘no education’ and ‘primary education’; Category 2 (medium education), which groups together the original categories referring to lower secondary education, upper secondary education, as well as professional post-secondary education; Category 3 (high education), which groups together the original categories referring to bachelor’s, master’s and PhD.
- 6.
The reference categories (baseline categories) are male (for ‘gender’), no education (for ‘educational attainment’) and married (for ‘marital status’).
- 7.
The AIC constantly decreased from the most basic model to the last model (AIC = 365,000 for the model that includes only gender, to AIC = 331,000 for the model that includes gender, age, education and marital status). The Log-likelihood ratio test was significant in each subsequent model, showing that each further model was better than the previous one in predicting the outcome variable.
- 8.
The goodness of fit indexes included in Table 7.2 are the AIC (Akaike Information Criterion) and two pseudo-R2. Together with the BIC, the AIC (Akaike Information Criterion) is one of the most used methods of model section (models with a lower AIC indicate a better fit). Pseudo-R2 is another way to assess the goodness of fit of logistic regression models (the higher the value, the better the model fits the dataset). To evaluate the best model among the estimated ones, I mostly refer to the AIC, because it is a measure of fit that favours more parsimonious models over more complex models (Field et al., 2012).
- 9.
Since some combinations between the clusters of the three typologies are not found in the data, in order to estimate the models 6, 7, 8 and 9, I created a new variable whose categories represent only the three-way interactions that are actually present in the data.
- 10.
The odds are defined as the probability that an event will occur divided by the probability that the event will not occur. Odds ratios are estimated in a logistic regression setting and vary between zero and +∞. If an odds ratio equals zero, it means that the outcome under observation (e.g. being a migrant worker in the domestic sector, instead of in another sector, for a woman instead of a man) is impossible to observe. When an odds ratio equals 1, then it means that the outcome is as likely to occur as its opposite (e.g. as migrants, women have the same probability as men to be working in the domestic sector or in another sector). In case the odds ratio is greater than 1, then the outcome being observed is more likely than its opposite (e.g. as migrants, women have a higher probability than men to work in the domestic sector instead of in another sector).
- 11.
This interaction corresponds to (1) Care 1: care regimes characterised by very strong familialisation and relatively generous care policies; (2) Gender 2: high scores in both dimensions of gender regimes; (3) Migration 3: migration regimes characterised by medium integration and open policies.
- 12.
The predicted probabilities correspond to the probability that an event occurs. Because they express absolute values, rather than relative values (as the odds ratio), predicted probabilities may be easier to interpret and are often preferred in social sciences to illustrate the results of logistic regressions. The values of predicted probabilities always range from 0 to 1.
- 13.
Interactions ‘Care1 × Gender3 × Migration3’, ‘Care1 × Gender4 × Migration5’, and ‘Care1 × Gender3 × Migration4’.
References
Ambrosini, M. (2013). Immigrazione irregolare e welfare invisibile. Il lavoro di cura attraverso le frontiere. Il Mulino.
Bryan, M. L., & Jenkins, S. P. (2015). Multilevel modelling of country-effects: A cautionary tale. European Sociological Review, 32(1), 3–22.
Cox, R. (2006). The servant problem. Domestic employment in a global economy. I. B. Tauris.
Davies, K. (1994). The tensions between process time and clock time in care-work: The example of day nurseries. Time & Society, 3(3), 277–303.
Field, A., Miles, J., & Field, Z. (2012). Discovering statistics using R. Sage.
Gerhard, U., Knijn, T., & Weckwert, A. (2005). Working mothers in Europe. A comparison of policies and practices. Edward Elgar.
Giordano, C. (2019). The Role of Gender Regimes in Defining the Dimension, the Functioning and the Workforce Composition of Paid Domestic Work. In Feminist Review, 122(2), 95–117.
Goñalons-Pons, P. (2015). Modern domesticity: Why professional women hire domestic workers in Spain. In A. Triandafyllidou & S. Marchetti (Eds.), Employers, agencies and immigration. Paying for care. Ashgate.
Hox, J. (1998). Multilevel modelling: When and why. In I. Balderjahn et al. (Eds.), Classification, data analysis and data highways. Springer-Verlag Berlin.
Lutz, H. (2008). Migration and Domestic Work. A European perspective on a global theme. Farnham, UK: Ashgate.
Lutz, H. (2011). The new maids. Transnational women and the care economy. Zed Books.
Maas, C., & Hox, J. (2005). Sufficient sample sizes for multilevel modelling. Methodology, 1(3), 86–92.
Saraceno, C. (2011). Childcare needs and childcare policies: A multidimensional issue. Current Sociology, 59(1), 78–96.
Saraceno, C., & Keck, W. (2010). Can we identify intergenerational policy regimes in Europe? European Societies, 12(5), 675–696.
Sarti, R. (2008). The globalisation of domestic service – An historical perspective. In H. Lutz (Ed.), Migration and domestic work: A European perspective on a global theme. Ashgate Publishing.
Twisk, J. (2006). Applied multilevel analysis. Cambridge University Press.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
Copyright information
© 2022 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this chapter
Cite this chapter
Giordano, C. (2022). Measuring the Impact of Care, Gender and Migration Regimes on Migrant Domestic Work. In: Ethnicisation and Domesticisation. Migration, Diasporas and Citizenship. Palgrave Macmillan, Cham. https://doi.org/10.1007/978-3-031-16041-7_7
Download citation
DOI: https://doi.org/10.1007/978-3-031-16041-7_7
Published:
Publisher Name: Palgrave Macmillan, Cham
Print ISBN: 978-3-031-16040-0
Online ISBN: 978-3-031-16041-7
eBook Packages: Social SciencesSocial Sciences (R0)