Considerations on the Environmental and Social Sustainability of Animal-Based Policies

: The contribution of the livestock sector to greenhouse gas (GHG) emissions as well as the worsening of animal welfare, with the intensiﬁcation of production methods, have become increasingly relevant. Our contribution investigates the environmental impacts, in terms of methane and nitrous oxide emissions, of animal-based policies supported by the European Union. We examine factors a ﬀ ecting the adoption and the magnitude of related budget of Measure 215—animal welfare—of Rural Development Programmes 2007–2013. Our focus is cattle farming in Italy. The results highlight that the problem of animal welfare is highly perceived in regions with greater livestock intensity, also where GHG emissions are relevant. Given the adoption of measure 215, more budget tends to be allocated in regions where livestock units are particularly high. In addition, from the analysis emerges the bargaining position of regions with a higher propensity to the agricultural sector.


Introduction
The sustainability of the agricultural practices has been longstanding at the forefront of the policy agenda in the European Union (EU). In fact, agriculture, while sustaining food production, is responsible for great impacts on the environment [1]: A remarkable example is the livestock sector [2,3]. Since the 1992 reform, the Common Agricultural Policy (CAP) paid special attention to the improvement of environmental performances of farmers; the mid-term review of the CAP in 2003 emphasised farmers' responsibility to adhere to a set of environmental, food quality and animal welfare standards. The greening of the direct payments, introduced in the EU greening reform in 2013, stimulated the transition towards more sustainable systems [3]. In addition, the greening architecture of the CAP 2014-2020, which reinforced mandatory and voluntary measures of cross-compliance, reflects the increasing political interest in the environmental implications of agricultural practices. In particular, the Council Regulation (EC) No. 73/2009 ratifies a mandatory 'cross-compliance' (Chapter I), based on which a farmer receives direct payments conditional to the compliance with Statutory Management Requirements (Article 5, Annex II) and Good Agricultural and Environmental Condition (Article 6, Annex III). One of the Statutory Management Requirements is animal welfare (Article 5(1c)), laying down minimum standards for the protection of animals that Member States have to achieve. Currently, the cross-compliance is regulated by the Regulation (EU) 1306/2013 (Article 93).
In such a framework, the livestock sector is receiving high pressures to address sustainability challenges, both from environmental and social perspectives [4]: In fact, the contribution of livestock farming to greenhouse gas (GHG) emissions and to worsening of animal welfare have become topics of utmost importance. Table 1. A synthetic outline of the literature on the environmental and social issues related to the livestock sector.

References Topic Subject
Gaudino et al. [3] Livestock sector as relevant contributor of GHG emissions Cattle, Italy Pelletier et al. [4] Methods to improve environmental/social sustainability of livestock Poultry, Canada Vitousek et al. [5] Correlation between livestock density and GHG emissions Livestock Steinfeld et al. [7] Livestock sector as relevant contributor of GHG emissions Livestock da Silva et al. [9] Livestock intensification and environmental/social concerns Livestock, Brazil Cederberg and Stadig [14] GHG emissions of livestock higher than other agricultural sectors Cattle, Sweden Casey and Holden [15] GHG emissions of livestock higher than other agricultural sectors Cattle, Ireland Lovett et al. [16] GHG emissions of livestock higher than other agricultural sectors Cattle, Ireland Pelletier et al. [18] Methods to improve environmental sustainability of livestock Poultry, United States Tallentire et al. [19] Methods to improve environmental sustainability of livestock Poultry Tallentire et al. [20] Methods to improve environmental sustainability of livestock Poultry Nieto et al. [21] Correlation between animal welfare and environmental impacts Cattle, Argentina Garnett [23] Correlation between livestock density and GHG emissions Livestock Havlík et al. [24] Livestock sector as relevant contributor of GHG emissions Livestock Feliciano et al. [25] Livestock sector as relevant contributor of GHG emissions Livestock, United Kingdom While a number of studies have investigated the implications of animal welfare provisions on the environmental performance of the livestock sector, to the best of our knowledge none have analysed which factors affect the adoption of measures of animal welfare. We aimed to fill this gap by investigating if site-specific characteristics are able to influence the definition of the political line at a regional level, with consequent impacts on the decision to adopt measures of animal welfare and the magnitude of the related budget. Our focus is on Measure 215 of Rural Development Programmes (RDP) 2007-2013, applied to cattle farms in Italy. Measure 215, regulated by the Council Regulation (EC) No. 1698/2005 (Articles 36(aV) and 40), provided annual subsidies for livestock farmers that voluntary adopted standards to ensure animal welfare higher than mandatory requirements listed in the Council Regulation (EC) No. 1782/2003 (Article 4, Annex III). Measure 215 compensated livestock farmers for incremental costs, or losses of income, of ensuring higher animal welfare levels (e.g., improvement of housing conditions) for a period ranging between five and seven years. We also quantified the environmental impacts, in terms of methane and nitrous oxide emissions, associated with the adoption of Measure 215 in Italy, by providing regional and provincial details. We comment on technical and political considerations, and conclude on how the measure may conciliate environmental and ethical issues. Our contribution would be of interest for policymakers that are planning to implement animal-based policies.

The Econometric Model
In order to investigate factors affecting the adoption and the magnitude of the budget allocated to measures of animal welfare (Measure 215), by regional Rural Development Programmes 2007-2013, we performed econometric analyses to assess how the characteristics specific for the i-th province in the j-th region influenced the decision to activate the Measure and the decision on budget allocation: where the dependent variable (Y ij ) is regressed on the province-specific livestock units (LU ij ), and utilised agricultural area (UAA ij ). We also controlled for K size classes of livestock (head k,ij ). β, γ, and δ k are the parameters to be estimated; α is a constant and ε ij an i.i.d. error term. Equation (1) is estimated through a Probit model to address the research question on the decision to activate the Measure, and through a Tobit model to address the decision on budget allocation. We use the Probit model to identify the drivers affecting the probability of implementing measures of animal welfare (i.e., extensive margins of Measure 215). In addition, the Tobit model allowed us to investigate which factors influenced the magnitude of the budget allocated to animal welfare in regions where the measure was implemented (i.e., intensive margins of Measure 215).
In other words, the Probit model allowed us to understand which factors influence the probability that the measure of animal welfare is adopted (or not) in a certain region; given the adoption of the measure of animal welfare, the Tobit model helps in analysing how the budget for measure of animal welfare is allocated across regions ( Figure 1).
where the dependent variable ( ) is regressed on the province-specific livestock units ( ), and utilised agricultural area ( ). We also controlled for size classes of livestock (ℎ , ). , , and are the parameters to be estimated; is a constant and an i.i.d. error term. Equation (1) is estimated through a Probit model to address the research question on the decision to activate the Measure, and through a Tobit model to address the decision on budget allocation. We use the Probit model to identify the drivers affecting the probability of implementing measures of animal welfare (i.e., extensive margins of Measure 215). In addition, the Tobit model allowed us to investigate which factors influenced the magnitude of the budget allocated to animal welfare in regions where the measure was implemented (i.e., intensive margins of Measure 215).
In other words, the Probit model allowed us to understand which factors influence the probability that the measure of animal welfare is adopted (or not) in a certain region; given the adoption of the measure of animal welfare, the Tobit model helps in analysing how the budget for measure of animal welfare is allocated across regions ( Figure 1).  The dependent variable (Y ij ) is a dummy variable (Measure active or not) in the Probit model and a continuous left-censored variable (relative budget for the Measure, compared to the budget for the regional Rural Development Program) in the Tobit model. The dummy variable equals 1 if Measure 215 was implemented in the j-th region, and 0 otherwise. The dependent variable of the Tobit model equals the incidence of financial resources allocated for Measure 215 on the total budget of the European Agricultural Fund for Rural Development (EAFRD) in the j-th region.
The Probit model is estimated controlling for the work intensity for livestock, expressed as the ratio between livestock unit (LU) and annual working unit (AWU), and the number of farms that adopted organic production methods. The rationale is that the greater the work intensity for livestock, the more the attention to animal welfare. Similarly, provinces that attach more attention to organic production methods tend to be open to animal welfare issues.
We did not introduce these two control factors in the Tobit model: They are likely to be correlated to the probability that Measure 215 was adopted (or not) in a certain region, but they were not able to influence the relative budget for Measure 215.
In order to refine our models, we estimate five specifications using as independent variables different statistics of the distributions of LUs and UAA. The model using total LUs and UAA allowed us to have an overall assessment of factors affecting the adoption and the magnitude of the budget for Measure 215. Minimum, maximum, and average LUs and UAA, as well as the difference between maximum and minimum LUs and UAA help the understanding of the influence of province-specific characteristics in the definition of the political line at a regional level. It should be noted that the use of extreme statistics is necessary in that while dependent variables are regional-specific, the variability is provided by different values at the provincial level. Put differently, the model must capture the "within" variation (where the term "within" is specific of panel data econometric models). Table 2 lists and describes the dependent and independent variables. Source: elaboration on Istat [26], BDN [27], and Rete Rurale Nazionale [28]. Notes: descriptive statistics refer to the Italian provinces in 2010.
We used cross-sectional data referring to 2010, due to the limited availability of data at regional and provincial levels. We collected information on 20 Italian regions and 107 provinces. Data were from the Italian database of the livestock registry (BDN) for LUs and size classes of livestock, from the 6th General Agricultural Census [26] for UAA, LU/AWU, and the number of farms with organic production methods. The official website of Rete Rurale Nazionale provides information on regional Rural Development Programmes (RDPs) from 2007-2013, on measures of animal welfare and the relative budget.

A Quantification of the Environmental Impacts
The livestock sector exerts a large influence on global greenhouse gas (GHG) emissions [25]: Livestock-related emissions mainly concern methane (CH 4 ) and nitrous oxide (N 2 O) and, to a lower extent, carbon dioxide (CO 2 ). CH 4 emissions are a by-product of enteric fermentation and of manure management; N 2 O emissions come from manure management, and from manure applied to soil or left on pasture [29].
The higher the intensity of livestock units (LUs) per hectare of utilised agricultural area (UAA), the greater the environmental impacts in terms of CH 4 and N 2 O emissions: where z indexes alternatively emissions of CH 4 and N 2 O; the environmental impacts (EI z ij ) depend on the livestock intensity (LU ij UAA −1 ij ) in the i-th province in the j-th region, and on the implied emission factor (EF z ) specific for CH 4 and N 2 O; α is a constant measuring the degree of improvement of livestock intensity (in percentage) due to the adoption of Measure 215. Data on EF z values depend on the region and the country average annual temperature. They are country-specific, refer to 2010, and are expressed in kg CH 4 head −1 for CH 4  Animal-based policies affecting livestock intensity are likely to influence both the levels of animal welfare and of GHG emissions. In fact, an improvement in livestock intensity may be achieved through a reduction of the number of LUs, being equal the UAA. Such a strategy is in line with the European Convention for the Protection of Animals kept for Farming Purposes, which outlines general principles intended to avoid unnecessary pain, suffering or injury due to unsustainable housing, environmental, and feeding conditions. But an approach based on the reduction of livestock numbers may also be beneficial for the sustainability of the environment. In fact, as suggested in Garnett [23], the lower the intensity of livestock, the higher the GHG saving: The trade-off consists in outweighing high emissions per unit through a significant reduction in numbers of units.
Accordingly, we assume that the environmental impacts (EI z ij ) depend on the degree of improvement of livestock intensity, achievable through a progressive, linear decrease (α) in the ratio LU ij UAA −1 ij . We quantified these effects by measuring how the environmental impacts vary according to different levels of improvement in livestock intensity. In particular, we considered five scenarios and assessed the environmental impacts at a national level related to a progressive decrease of the livestock intensity, with α ranging between −10% and −50%.

Regression Results
The results of Probit and Tobit estimations are in Tables 3 and 4 respectively. Each model considers five specifications according to the type of independent variables: total livestock units (LUs) and utilised agricultural area (UAA) (columns A), minimum (columns B), maximum (columns C), and average (columns D) LUs and UAA for each region, difference between maximum and minimum LUs and UAA in each region (columns E).
The results suggest that the higher the LUs in the province with the minimum LUs in a certain region, the higher the likelihood of adopting a measure of animal welfare in that region. In addition, it was more likely that a measure of animal welfare was adopted in regions where UAA in the province, with the maximum UAA in a certain region, or the region-specific difference between the provinces with the minimum and maximum UAA tended to be low. The minimum LUs and maximum UAA are proxies of the intensity of LUs per hectare of UAA in a region: If LUs are high in the province with the minimum LUs in a region, or UAA was low in the province with the maximum UAA in a region, the intensity of LUs per hectare of UAA was more relevant for that region.
The rationale is that the problem of animal welfare is highly perceived in regions with great numbers of LUs or with limited UAA. In other words, a measure of animal welfare tends to be adopted in regions with high intensity of LUs per hectare of UAA.
In addition, a positive relation links the adoption of measures of animal welfare with both 'LU/AWU' and 'Organic'. Provinces that attached more attention to organic production methods tended to be open to animal welfare issues. The results also reveal that the greater the work intensity for livestock (LU/AWU), the more the attention to animal welfare. Robust standard errors are in parentheses. ***, **, and * indicate statistical significance at 1%, 5%, and 10%. Notes: Coefficients are on the order of 10 −3 . Acronyms are livestock unit (LU), utilised agricultural area (UAA), annual working unit (AWU).
As for the analysis of factors affecting the magnitude of the budget allocated to measures of animal welfare (Measure 215) by regional Rural Development Programmes 2007-2013 (Table 4), positive significant coefficients were estimated for 'UAA' (0.00059), 'UAA (min)' (0.00108), and 'UAA (mean)' (0.00142). Similarly we found a positive significant coefficient for 'LU (min)' (0.00715). As for size classes of livestock, positive (negative) significant coefficients were estimated for '100-499 heads' ('More than 500 heads') in one out of five model specifications (Table 4, column B).
Given the adoption of the measure of animal welfare, more budget tended to be allocated in regions where LUs were particularly high, that is, LUs were relevant in the province with the minimum LUs in a certain region. It is not surprising that regions where the problem of animal welfare is emphasised (i.e., high intensity of LUs) need more financial resources to face the problem.
Similarly, having a great UAA contributed to increasing the budget allocated for animal welfare. This was particularly true in regions where UAA was high in the province with the minimum UAA. This evidence highlights the bargaining position of regions with higher propensity to the agricultural sector. Robust standard errors are in parentheses. ***, **, and * indicate statistical significance at 1%, 5%, and 10%. Notes: Coefficients are of the order of 10 −3 . Acronyms are livestock unit (LU), utilised agricultural area (UAA).

Environmental Impacts Assessment
The assessment of environmental impacts, in terms of methane (CH 4 ) and nitrous oxide (N 2 O) emissions, related to cattle in each Italian region is in Table 5 [28].
In order to examine the extent to which animal-based policies were able to affect the environment, we quantified the impacts, in terms of CH 4 and N 2 O emissions, associated with different improved levels of animal welfare (Table 6). For the sake of argument, we considered five scenarios and assessed the environmental impacts at a national level related to a progressive decrease of the livestock intensity, with α ranging between −10% and −50%. Accordingly, and on the basis of equation (2), the reduction of livestock units (LUs) being equal the utilised agricultural area (UAA) reduced CH 4 and N 2 O emissions: The greater the improvement of livestock intensity, the higher the greenhouse gas (GHG) saving. Source: elaboration on Istat [26], BDN [27], Rete Rurale Nazionale [28], and Faostat [30]. Notes: elaborations refer to 2010. However, the environmental impacts were larger with lower improvements of livestock intensity: For instance, the impacts were twice lower moving from −10% to −20% of livestock intensity, but only 1.3 times lower from −30% to −40% or from −40% to −50% (Table 6). Consider a livestock intensity equal to 0.1 head ha −1 (say 10 heads per 100 hectares of UAA), a 10% reduction in livestock intensity (say 9 heads per 100 hectares of UAA) would reduce methane emissions by 6.821 kg CH 4 ha −1 and nitrous oxide emissions by 0.002 kg N 2 O-N kg N per head ha −1 .

Discussion and Conclusions
Our analysis is an ex post assessment of animal-based policies, such as Measure 215 of the Rural Development Programmes (RDP) 2007-2013, intended to ensure higher levels of animal welfare.
The results highlighted that the problem of animal welfare is highly perceived in regions with high density of livestock: thus it is more likely that animal-based policies are adopted in regions with greater environmental impacts.
We also found that, once Measure 215 had been adopted, financial resources tended to be allocated in regions where the problem of animal welfare is emphasised or where the high propensity to the agricultural sector drives the bargaining position of those regions.
It is worth mentioning that, in our sample, Measure 215 tended to be adopted by regions with particularly relevant environmental impacts, such as Veneto, Campania, Piedmont, and Aosta Valley. Our results are in line with Pereira and Martinho [31], who argued that the emission performance of policies is highly differentiated and site-specific. Information on agricultural emissions at the regional level may be important indicators to assess whether the demands for emissions reduction targets, set by government, are likely to be met [25,32].
Measure 215 is thought to ensure higher levels of animal welfare preserving efficiency and quality of animal-based production. In particular, the measure gives specific attention to the development in housing conditions: Iif implemented, Measure 215 aims to ensure higher housing areas. At the farm level, a potential strategy to improve the housing areas is to reduce the number of heads, leaving farm sizes unchanged. However, such a strategy has implications on the environment that should not be neglected: In fact, livestock are major contributors of methane and nitrous oxide emissions [7].
A policy measure, that aims to increase the housing areas by reducing the number of heads while keeping constant the size of the farms, allows a reduction of environmental impacts in terms of methane and nitrous oxide emissions. In fact, our analysis shows the reduction of livestock numbers lead to genuine GHG benefits, as also suggested in Garnett [20]. We found coherence between animal-based policies and other greening requirements of the Common Agricultural Policy (CAP). As suggested in Solazzo and Pierangeli [33], greening measures have limited environmental impacts. In particular, in line with previous literature, we found that Measure 215 contributed to reduce nitrogen [34] and greenhouse gas emissions [35].
Given the increasing environmental and ethical concerns related to the livestock sector, policymakers should enable regional-scale strategies and include specific voluntary measures in their national planning context to contribute to the improvement of animal welfare, as well as to the reduction of GHG emissions.