**3. Results and Discussion**

The descriptors and the regression coefficients of the model are presented in Table 3. Together, the independent variables are statistically significant in estimating the water footprint (*p* < 0.00). According to the R squared statistic, 43% of the total variation of WFP is explained by the model. The model was also checked for multicollinearity as mentioned above. The variance inflation factor (VIF) value obtained was close to one and, thus, there was no evidence of multicollinearity [48]. To evaluate the relative importance of the independent variables, it is common to calculate the beta coefficients (standardized regression coefficients). In a regression of standardized variables, the (beta) coefficient estimates express the rank of independent variables in terms of the effect on the dependent variable. The independent variable with the largest (absolute) beta coefficient has the biggest effect on the dependent variable. The intercept in such a regression is zero by construction. According to the results, the F-ratio test confirms that the overall regression model is a good fit for the data (Table 3). The output shows that the independent variables statistically significantly predict the dependent variable.


**Table 3.** Results of the semi-log multiple linear regression model (n = 4853).

\*, \*\*, and \*\*\*, statistically significant at 10%, 5%, and 1%, respectively.

According to Table 3, both demographic variables (household size and age of the head of the household) are very significant (at 1%) in the prediction of the water footprint of food consumption of the household with coefficients −0.044 and 0.002, respectively. The city/geographic variable it is statistically significant at the 5% level; people with higher water footprints are more likely to be found in big cities than in small and medium ones. Region is very significant, with only the southeast not differing from the base of Tunis. The centre-east, centre-west and northeast have the highest coefficients: 0.165, 0.155, and 0.116, respectively. Concerning the socio-economic variables, poverty (−0.221), education particularly illiterate people (−0.040), and the socio-professional categories (SPC), especially farmers (0.108), are also very significant. Finally, variables related to food habits (Expenditures and Food waste) are also significant at the 1% level.

In terms of the relative importance of the effects on the dependent variable, based on beta coefficients, food expenditure per capita (0.456), household size (−0.157), and poverty (−0.151) have the largest contributions across the model. We also find that the centre-east (0.125) and centre-west (0.114) regions have the largest effects on the water footprint. This is followed by food waste, represented by the number of dishes thrown away with a beta coefficient equal to 0.069, the socio-professional category "farmer" (0.064), the age of household head (0.056), the education level of the household head, and, finally, the variable "City size" determining the size of the city of residence. The City size variable is linked to the degree of the economic development of the city. According to Souissi et al. [18], the evolution and increase in water footprint during the last thirty years in Tunisia is more rapid in urban regions. The more developed the city is and the better the economic situation, the higher the household water footprint. A 1 TND (US\$ 0.69) increase in food expenditure is associated with 0.04% increase in the average water footprint. This can be explained by the increased consumption of animal products, which are usually more expensive than plant products [49]. Meat and dairy products have a significant impact on the water footprint. This is an alarming sign, especially since the measured footprint is mainly internal (more than 70% of the water footprint of the main food products comes from local production) [18]. In other words, Tunisia is severely depleted of internal water resources by consumption habits.

A one-unit increase in the size of the household implies a 4% decrease in the average food consumption water footprint, controlling for food expenditure. Poor households have a 22% lower water footprint than other households. Wealthier households seem to consume products with a large water footprint.

Region is also an important factor to determine the water footprint of households. The average water footprint is, respectively, higher by 16%, 15%, 11%, and 8% for households living in the centre-east, the centre-west, the northeast, and the southwest of the country than for people living in Tunis. The centre-east and northeast regions are characterized by high economic development and tourism. Households' incomes are higher and access to more expensive food products, especially of animal origin, is better. Concerning the centrewest and the southwest, these regions are characterised by sheep and goat production, resulting in meat being both available and culturally important. Meat consumption is the highest in the southwest of the country. The average water footprint for people living in the northwest is 6% lower than for people living in Tunis. The diet in the northwest is based on cereal products, which has a lower water footprint. This region is less economically developed and has substantial cereal production. There is no significant difference between the water footprint for households living in the southeast and those living in Tunis. These results can be explained by the variation in culinary habits from one region to another. Regional food patterns are often very pronounced in Tunisia, particularly for meats [49].

Regarding food waste, all other variables being constant, we found that for each dish thrown away by the household the water footprint increases by 1.5%. Li et al. [33] found similar results showing that the increase in food waste contributes to a higher water footprint. For the socio-professional categories of the head of the household, the average water footprint is, respectively, higher by 10%, 5%, 4%, and 3% for farmers, freelance jobs, industry and commerce independents, and employees than for labourers.

Considering the effect of the head of the household's age, the unstandardized coefficient for the variable age is equal to 0.002. This means that for each one-year increase in the age, there is an increase in the average water footprint of 0.2%, all other variables held constant. It is hard to explain this small but very significant effect of age on the water footprint. On one hand, the increase with age may imply the presence of children, whose food consumption is characterised by incorporation of dairy products, meats, and cold cuts [50]. On the other hand, studies in other countries have shown that the oldest consumers ate more vegetables and fruits as well as less meat and fewer sugary desserts [50,51]. For education, the average water footprint is 4% lower for illiterate heads of households than for those with primary education. There is no significant difference between the other categories. Finally, regarding the city size, results show that the average water footprint is 3% lower in medium and small cities than in big cities. The effect of urbanization should not be overlooked. Urbanization was involved in our analyses due to the association of urbanization and the structure of the diet in many studies [52–56]. The literature examined shows that, unlike rural diets, urban diets are more characterized by the consumption of flour, more fat and animal products, more processed food, more sugar, and more food consumed outside the home. All of these elements necessarily impact the water footprint, which continues to climb in urban areas.

#### **4. Conclusions and Policy Implications**

The determinants of a consumer's water footprint depend on the water footprint of the goods produced. It also depends on what the consumer chooses to consume and the consumed quantities. Until now, studies related to the water footprint have not highlighted the factors affecting these choices nor their contributions to the water footprint of consumers.

In this paper, to better understand the factors that influence the food water footprint of Tunisian consumers, we used a semi-log multiple regression model. Results show that the increased consumption of animal origin products is necessarily linked to the increase in food expenditure per household and has a significant role in the water footprint increase. Demographic and economic characteristics such as household size and poverty are among the factors that contribute to the decrease in the consumer's water footprint. Moreover, regional disparities in food choices mean substantial differences in water footprints. Residents of the most developed cities and coastal cities in the centre-east, centre-west and northeast are more likely to have a large water footprint than residents of Tunis. Significant variability in water footprints, due to the different modes of food consumption and sociodemographic characteristics, was also noted. Food waste is one of the determining factors of households with a large water footprint.

This study contributes to the literature on the water footprint of food consumption using household level data. Estimates of the food water footprint can be used to assess potential scenarios for water demand as food consumption patterns change. Reducing the water footprint to sustainable levels is possible if consumption patterns change.

Analysis at geographic and social levels helps inform policy makers by identifying realistic dietary changes, taking into account socio-economic and regional disparities to effectively plan interventions and recommendations for a sustainable diet. It would be important to encourage more sustainable diets rich in vegetables and fruits, in particular through schools and advertising campaigns. In addition, in accordance with sustainable development goals and, in particular, objectives two (SDG2), six (SDG6), and twelve (SDG12), namely, to end hunger, ensure availability, and sustainable management of water and reduce food waste, it will be necessary to reconsider import and export strategies for food/agricultural products as well as food subsidy policies. For example, the wheat import strategy is effective during years when world prices for cereal products are lower than the cost of production. This allows Tunisia to save very important volumes of water. However, for reasons of food security and food sovereignty, the cultivation of wheat should

be encouraged especially in more humid areas, especially in the north-west where the diet depends mainly on these products.

Several economic and political mechanisms aimed at reducing the water footprint of food consumption are possible. On the one hand, this may be achieved by relying on supply chain marketing strategies such as labeling. On the other hand, on an international scale, the ISO 14046 standard specifying the principles, requirements, and directives relating to the evaluation of the water footprint of products and processes has been established. Other measures based on food price and subsidy policies as well as consumer awareness campaigns can yield tangible results. Agricultural policies can also be an effective tool to reduce the water footprint of food consumption.

However, conclusions and recommendations should be viewed with caution since several limitations are noted in the use of this concept. The main limitations are the imprecision of the estimates, which is due to the difficulty of estimating water consumption at all stages of the food chain. Water volumes for products vary depending on production systems, rainfall, soil quality, yields, irrigation, etc. Other factors affect other aspects of the food chain, so imprecision accumulates. In addition, only the main food groups are considered and the data do not include fish products. In addition, the insufficiency of the volumetric approach should not be overlooked, since in addition to the volume of water consumed, the quality and conditions of access to water also play a role in decisionmaking regarding the use of resources. Another difficulty is the evaluation of grey water; determining the volumes of water "hypothetically" necessary to dilute the pollution to a tolerable level is quite arbitrary and very complex. To conclude, we can say that the use of the water footprint must take into account several limits depending on the context and the objective.

Finally, the absence of previous work that models the factors influencing the water footprint of food consumption opens up several perspectives for future research. The exploration and identification of new influencing variables (such as diet diversity, processed food consumption, etc.) and the use of more recent data that take into account postrevolutionary political and social changes in Tunisia are a priority.

**Author Contributions:** Conceptualization, A.S. and N.M.; methodology, L.M., N.M. and C.T.; software, A.S.; validation, L.M., N.M., C.T. and A.C.; formal analysis, A.S. and N.M.; investigation, A.S.; resources, L.M.; data curation, A.S.; writing—original draft preparation, A.S. and N.M.; writing review and editing, L.M., N.M. and A.C.; visualization, A.S.; supervision, C.T. and N.M.; All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Appendix A**

**Table A1.** Akaike's information criterion and Bayesian information criterion.


Note: N = Obs used in calculating BIC.

#### **Appendix B**

estat hettest, rhs Breush–Pagan/Cook–Weisberg test for heteroscedasticity Ho: Constant variance

Variables: AgeChefMe Nombredeplatsjetes vuln taille DAP 1b.newstrate 2.newstarte 1.DNiveau 2b.DNiveau 3.DNiveau 4.DNiveau 1b.region 2.region 3.region 4.region 5.region 6.region 7.region 1.DCSP 2.DCSP 3.DCSP 4.DCSP 5b.DCSP 6.DCSP 7.DCSP Chi2(21) = 467.20 Prob > chi2 = 0.0000
