The aim of the present study was to estimate the prevalence of food insecurity in the Paris area, using a standardised instrument that provides more precise indications and a more precise definition of food insecurity (and of its severity), thanks to the collection of declarative data on a set of lived experiences. Our study was the first one to have used this instrument in France, and it had the attribute of being population-based and representative of the Paris metropolitan area.
We found an overall FI prevalence of 6.30%, with a prevalence of very low FS of 2.40%. In this study, we identified several characteristics of HHs that had experienced FI within the previous 12 months. We found a higher prevalence of FI among HHs receiving welfare, two-or-more-family HHs and, of course, the poorest HHs. On the other hand, the prevalence was particularly low among HHs headed by an upper-white-collar worker or an elderly person.
Once adjusted for income, certain determinants were significantly associated with FI. First, the effect of sociodemographic variables (HH head age, gender and education level), economic indicators (source of income and residing in social housing) and contextual variables that was observed in univariate analysis disappeared after adjustment for income, which may be evidence of its strong impact. Second, the presence of a child under 3 years of age in the HH and being a single-parent family both remained associated with FI in the poorest group. This raises questions in the French context, where one might expect welfare policies to be specifically directed at single-parent families and at preventing children from experiencing food insecurity. Third, in the two groups above the poverty threshold, HH family composition was the most important determinant, after adjustment for income level. Indeed, when compared to single-family HHs, all other types of HHs were at higher risk for FI, which might be due to the additional socioeconomic constraints experienced by these families.
These results have some limitations. First, because of this study’s cross-sectional design, we cannot conclude that a particular family composition causes food insecurity, but rather that it only describes certain family types that are especially at risk. Second, it should be mentioned that our version of the HFSSM questionnaire contained fewer questions regarding child FI. However, this had no impact on our estimate, since we analysed FI among adults only and separately, as did a Canadian study previously . One limitation of this study is the sample size, which may have sometimes resulted in a lack of strength in our analysis, but the punctual estimate (i.e., the OR estimate) can give an indication of the kind of association observed. A limitation of this study lies on the fact that our sample excluded homeless (population estimated in 2010 at almost 21 200 in Paris metropolitan area ) or non-French speaking people, who are certainly a population that may experience food insecurity. These exclusions may induce an under-estimation of the food insecurity phenomenon. Another limitation (common to all surveys that use this tool) is the failure to investigate all the dimensions of food insecurity as originally defined (particularly the social acceptability of food acquisition). Also, response bias may have resulted from the shame attached to FI, which may therefore have been underreported by the interviewees and have led to an underestimation of the prevalence of FI. Lastly, only one person in the HH answered the questionnaire, which was used to estimate FI for the entire HH. However, since 99% of the respondents were the HH head or his/her partner, they had a good knowledge of the situation in the HH. While the use of the HFSSM has been discussed in several studies and reports [34, 35], we think that it is particularly useful for describing the situation in France because it enables one to estimate different ranges of FI severity and make international comparisons.
Two other indicators of FS (see Additional file 2) had been used previously in a secondary analysis of the Individual and National Food Consumption Survey (INCA 2)c and in the Health and Nutrition Barometer (BSN)d . In our survey, the FI prevalence was much lower than the prevalence of food insufficiency, as it can be estimated by applying the INCA 2 and BSN instruments to our study population: 6.30% for FI versus a food insufficiency prevalence of 20.9% and 10.5% in the BSN and INCA 2, respectively. This difference can be explained by the conceptual differences captured by these tools, for the HFSSM has a narrower definition of food hardship than the other two instruments.
The prevalence rate found in our study also seemed to be lower than the rates estimated with the same questionnaire in other Western countries. For example, the FI prevalence was found to be 14.5% in the U.S. in 2010  and 7.7% in Canada in 2007–2008 . Of course, these national prevalence rates were not directly comparable with our results for the Paris metropolitan area. However, similar or even greater differences were observed when we compared our results with FI prevalence rates for the main cities in the U.S. (according to 2003 Office of Management and Budget delineation), where the annual American food insecurity study reported an FI prevalence of 17% (with 10.7% of the urban population experiencing low food security and 6.3% very low food security) . In the Montreal Health Region in 2008, 9.0% of HHs were living in FI (6.2% in low food security and 2.7% in very low food security) . The fact that the FI prevalence is so much lower in the Paris metropolitan area may be due to its socioeconomic characteristics, for this region is known to be the wealthiest in France, and Paris is the second wealthiest city in the European Union (EU), after London . But this may not be the only explanation, since the Paris region as a whole ranked 7th among EU regions in 2008 (as determined from the regional GDP per inhabitant by EUROSTAT) and is the region in France with the greatest social disparities . It may also be due to the national context of the French welfare state’s (still) generous safety net. In the mid-2000s, the OECD estimated that the income poverty rate, which is based on 50% of the median income after taxes and transfers was 7.2%, 11.4% and 17.0% in France, Canada and the USA, respectively .
Consistent with the findings of studies carried out in the USA [30, 42, 43], Canada  and England , our study found that the prevalence of FI increased as HH income decreased. Moreover, it confirmed that in France, too, household composition is a major factor for food insecurity, as it is in Canada  and in the U.S. , where single-family households are less food-insecure (especially when there are no children). In both of these countries [9, 30], households with children have a higher FI prevalence, and single-parent families are also more at risk, specifically, HHs with a single mother. This latter situation was observed in our study as well, in which single-parent families (most of which were headed by women) were at higher risk for FI in all income subgroups. In France, the specific family allowance for single parents ceased to exist as such and was included in a new, comprehensive allowance for the poor (which, in practice, is not very easy to obtain) right before the survey. As regards the association with education level, the gradient we observed was consistent with findings in the U.S.  but differed from the Canadian results , which did not seem to follow this pattern (perhaps because the proportion of the Canadian population with a tertiary education is particularly high).
Our results highlight some important points that may indicate specific vulnerabilities that characterise certain family situations and advocate for specific public policies targeted at these households (for example, special attention from a social worker). First, for the poorest households, we mentioned above the alarming situation regarding single-parent families and HHs with a preschool child. In France, in 2009, 4.5 million people were under the poverty threshold used in our models . A recent study described the effect of a birth on the HH standard of living  and showed how the existing child welfare benefits may only partially offset the expenses, especially in terms of the impact on the occupational activity - and income - of one of the adults in the HH (and the only adult in single-parent families). These findings are worrisome, given the literature on the impact on children of living in a food-insecure HH [48–51] and the fact that food-insecure adults may compromise their nutrition to maintain that of their children [6, 52]. Indeed, children’s nutrition, dietary intake and weight are affected, but this situation also has social, emotional and health (physical and mental) consequences [48–51, 53–55].
In our study, HH composition appeared to be a more differentiating factor than HH head gender. Although female-headed HHs had a higher prevalence of FI, in multivariate analysis, they were not at greater risk. Since we stratified across income categories and adjusted for income and family composition, our results suggest that the greater vulnerability of female-headed HHs may be due to their worse poverty or their specific family composition. Indeed, previous findings in the literature [9, 14, 56] point to single-mother households being at greater risk for FI. On the other hand, the greater vulnerability of single-parent families must be due to factors other than the parent’s gender or income (which were not taken into account in the models), such as the proportion of the HH budget available for food (after the other expenses, such as daycare, are paid) and the unavailability to shop, which means not being able to prepare food at a lower cost.
A significant finding of our study was the high prevalence of FI among two-or-more-family HHs. They may be at greater risk for FI because of the number of adults and, therefore, the large quantity of food required. In the Paris area, they are mostly immigrant households and consequently more likely to be in underprivileged and precarious conditions . Unfortunately, we could not test this hypothesis, since we did not know the immigration status of each member of the HHs. In the case of single-person and unrelated-persons households, we can assume, as we did for single-parent families, that the income variable used for our adjustment (before-tax total income per CU) conceals a lower disposable income than that of the reference (single-family) HH, once all the compulsory, nonreducible expenses are paid. A last result concerns the specific situation regarding the disabled, who were at greater risk for FI in all three income groups (even if the risk was significant only in the lowest group, with an extremely broad confidence interval). In France, the disabled have, on average, a particularly critical and low socioeconomic status  (their minimum net income guaranteed by Social Security is considerably below the poverty threshold).