Prevalence elasticities and socio-demographic characteristics by income groups
| Variables | Low-income households | Middle-income households | High-income households | All households |
| Prevalence elasticity (logit model) | ||||
| Price | −0.595*** (0.067) | −0.582*** (0.064) | −0.344*** (0.074) | −0.521*** (0.039) |
| Income | 0.328*** (0.052) | 0.255*** (0.069) | 0.296*** (0.060) | 0.325*** (0.025) |
| Household size | 0.077*** (0.029) | 0.160*** (0.049) | 0.037 (0.046) | 0.051*** (0.015) |
| Male ratio | 0.452*** (0.167) | 0.835*** (0.150) | 0.872*** (0.102) | 0.794*** (0.074) |
| Adult ratio | 0.529*** (0.171) | 0.606*** (0.207) | 0.328 (0.287) | 0.545*** (0.119) |
| Maximum education: more than secondary complete | ||||
| No education | 0.522 (0.329) | 0.252 (0.327) | 0.436 (0.273) | 0.407** (0.175) |
| Primary complete | 0.245** (0.118) | 0.395*** (0.122) | 0.321*** (0.115) | 0.304*** (0.067) |
| Secondary complete | 0.215** (0.089) | 0.230*** (0.082) | 0.255*** (0.079) | 0.232*** (0.047) |
| Region: centre | ||||
| South | −0.064 (0.117) | −0.221** (0.101) | −0.487*** (0.078) | −0.320*** (0.054) |
| North | −0.439*** (0.072) | −0.345*** (0.077) | −0.208** (0.099) | −0.379*** (0.046) |
| HH activity: employed | ||||
| Unemployed | 0.367*** (0.136) | 0.436* (0.236) | 0.031 (0.310) | 0.295*** (0.109) |
| Pensioners | −0.146 (0.098) | −0.188** (0.095) | −0.473*** (0.085) | −0.310*** (0.052) |
| Constant | −8.820*** (2.943) | −20.202*** (4.075) | −23.203*** (4.710) | −12.404*** (1.790) |
| Observations | 4172 | 4167 | 4164 | 12 503 |
Source: Own calculation; ***p<0.01, **p<0.05, *p<0.1.
Link test results showed that the linear predicted value squared is not statistically significant (p>z = 0.241). Hosmer and Lemeshow’s (HL) goodness of fit test indicates that our model fits the data well; the optimal number of groups of data used in the test is 1253, χ2=1385.41, prob >χ2=0.6531). Multicollinearity is not an issue—mean VIF is 1.45, and VIF income and income squared are 1.81 and 1.19, respectively. There is no statistically significant difference between price elasticities of low-income and middle-income groups (χ2(1)=0.08, prob >χ2=0.7712). A statistically significant difference between price elasticities of low- income and high-income group is confirmed (χ2(1)=6.13, prob >χ2=0.0033), as well as between medium-income and high-income groups (χ2(1)=7.53, prob >χ2=0.0061). The results showed the absence of statistically significant difference between income elasticities among all income groups: low-income and middle-income groups (χ2(1)=0.63, prob >χ2=0.4284); low-income and high-income groups (cχ2(1)=0.16, prob >χ2=0.6860); middle-income and high-income groups (χ2(1)=0.18, prob >χ2=0.6730). HH activity: If there is at least one member of the household who is employed, the household is defined as employed; If there are no employees, but there is a pensioner in the household, the household is defined as a pensioner; If the adult household members are all inactive or unemployed, the household is labelled as unemployed.