Territorial analysis of the diffusion of organic farming in France: Between heterogeneity and spatial dependence

G Model

ARTICLE IN PRESS

ECOIND-2343; No. of Pages 12

Ecological Indicators xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Ecological Indicators journal homepage: www.elsevier.com/locate/ecolind

Territorial analysis of the diffusion of organic farming in France: Between heterogeneity and spatial dependence Gilles Allaire a,∗ , Thomas Poméon a , Elise Maigné a , Eric Cahuzac a , Michel Simioni b , Yann Desjeux c a

INRA, US ODR, F-31326 Auzeville, France INRA/GREMAQ and IDEI, Toulouse School of Economics, F-31000 Toulouse, France c INRA, UMR 1302 SMART, F-35000 Rennes, France b

a r t i c l e

i n f o

Article history: Received 7 April 2014 Received in revised form 2 March 2015 Accepted 4 March 2015 Keywords: Organic farming Conversion Diffusion Spatial structure Spatial dependence Path dependence Collective capacities Agricultural policies

a b s t r a c t This paper discusses the development of organic farming (OF) in France from a collective point of view by focusing on the spatiotemporal diffusion of OF and considering different types of production. Based on the data on aid granted between 2007 and 2010 for conversion to OF (COF), the spatial dynamics of conversion is examined with regard to the distinctive capacities of micro-territories to accommodate farms engaged in OF to a greater or lesser extent. The hurdle model is applied to varying types of COF aid, which are related to different production systems. This allows for both the characterization of the geographical extent of the contracting of COF aid and its local intensity measured by the number of contracts within micro-territories. The spatial structure of COF contracting can be explained both by economic factors relating to the orientation of production systems and by phenomena of spatiotemporal dependence, which demonstrate the importance of producers’ experience and of collective capacities. We can therefore speak of path dependence in relation to the establishment and maintenance of market access capabilities and social networks, which determine the potential and effectiveness of the development of organic agriculture at the micro-territorial level. © 2015 Elsevier Ltd. All rights reserved.

1. Introduction This article provides a spatial analysis of the diffusion of organic farming (OF) in France, globally and according to the type of production. Previous work has shown both, that is, heterogeneity in the dynamic distribution of organic agriculture across regions and within them, and phenomena of local spatiotemporal dependence (Allaire et al., 2014, 2015). The originality of this study lies in the uniqueness of its approach and aim: separate analyses are conducted for different types of production, and in each case careful attention is paid to both the spatial extent of the phenomenon (i.e. the presence of at least one OF producer by spatial unit) and its intensity (number of producers) within each of the microterritories that constitute the spatial units observed. By addressing the development of OF from a territorial point of view, the focus shifts from individual adoption towards that of territorial diffusion. The spatial diffusion of OF is not just the result of the spatial distribution of factors influencing individual adoption

∗ Corresponding author. Tel.: +33 0561285086. E-mail address: [email protected] (G. Allaire).

of this technology. The territorial context (which includes social, economic, institutional and natural factors) plays its own important role in the diffusion of OF. Motivating or hindering forces that express themselves at different territorial levels partly explain the localization of organic farms. Certain locations are likely to accommodate OF farms (defining the extent of OF) and to accommodate them in greater or smaller numbers (defining the intensity of OF development), thereby making individual conversions more or less costly or risky. Several explanatory factors can be highlighted, such as the geography of production and of outlets (local markets and systems of collecting), the regional or local policies, and the local collective or institutional capacities. 1.1. The spatial distribution of organic farming Several studies have analyzed spatial differences in the diffusion of OF, based on concentration indices, at different levels and for different countries (Beauchesne and Bryant, 1999; Ilbery et al., 1999; Frederiksen and Langer, 2004; Eades and Brown, 2006; Risgaard et al., 2007; Ilbery and Maye, 2011; Allaire et al., 2014); or by modelling agglomeration and neighbourhood effects (Nyblom et al., 2003; Bichler et al., 2005; Gabriel et al., 2009; Lewis et al., 2011;

http://dx.doi.org/10.1016/j.ecolind.2015.03.009 1470-160X/© 2015 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Allaire, G., et al., Territorial analysis of the diffusion of organic farming in France: Between heterogeneity and spatial dependence. Ecol. Indicat. (2015), http://dx.doi.org/10.1016/j.ecolind.2015.03.009

G Model ECOIND-2343; No. of Pages 12

ARTICLE IN PRESS G. Allaire et al. / Ecological Indicators xxx (2015) xxx–xxx

2

Bjorkhaug and Blekesaune, 2013). More qualitative approaches have enabled illustrations of the diffusion logic at work at very local levels (Noe, 2004), or from a structural perspective of a regional sector (Boivin and Traversac, 2011). Most of this work, inspired by economic geography, does not limit itself to observations about the proportion of organic area in a territory, but implicitly assumes that it is the effect of an investment, or an effort, that may be the result of the addition of private or collective efforts and of various structural factors (natural, socio-economic, political, etc.). This underlying assumption is necessary for considering whether or not the localization of OF is due to chance alone. In an analysis of the spatial distribution of OF at the county level in Germany, Schmidtner et al. (2012) proposed two distinct and complementary spatial dynamics. They built on the proposal of Anselin (1988) by considering, first, the location factors that determine spatial heterogeneity (spatial structure) and, second, the agglomeration effects related to spatial dependence (also referred to as spatial autocorrelation or spatial interaction). Empirically, the authors used data aggregated at NUTS1 3 level (which is not a very fine resolution), due to the lack of availability of individual data. However, their reasoning was still based on the modelling of individual behaviour; explanatory variables related to factors that influenced the decision at the farm level. Agglomeration effects (of counties favouring OF) therefore corresponded to spillover effects. In order to separate these two diffusion dynamics, the present study focuses on the proper role of territorial institutional capacities in explaining observed phenomena of spatial and, more precisely, spatiotemporal dependence. The role of context cannot only be analyzed in terms of mere externalities, there are also increasing returns to adoption due to systemic economies of scale associated with the concentration of organic farms. These are partly territorialized and include formal and informal networks of organic farmers, technical support structures, downstream structuring, etc. Experience, individual and collective, acquired at the territorial level can improve the territory’s institutional quality. We can therefore speak of “path dependence” (Bichler et al., 2005; Allaire et al., 2014), which is reflected by spatial or temporal autocorrelation in diffusion models. This leads us to consider the heterogeneity and spatial dependence of OF diffusion beyond individual behaviours and determinants. The path dependence is both systemic (network effects on learning costs) and based on collective capabilities (clarification of opportunities) derived from collective past experience. These are the phenomena that are discussed here, starting with an analysis of public aid contracting for conversion to organic farming (COF) in France, during the period from 2007 to 2010, valorizing exhaustive administrative data. The article’s specific aim is to explore the causalities involved in OF geographic diffusion according to different types of production, based on an econometric model of the extent and intensity of contracting, and by taking the characteristics of micro-territories, as defined on the basis of the NUTS4 scale, into account. The proposed model is designed to distinguish spatial heterogeneity from spatial dependence. Before presenting the data we used, our models and our results, we assess the recent dynamics of OF diffusion in France. 1.2. The dynamics of organic farming diffusion in France Although French organic farming continues to lag, with a share of 3.1% of the UAA2 (utilized agricultural area) in 2010 (Agence Bio, 2012), it grew significantly between 2007 and 2010, with a

1

The NUTS classification (nomenclature of territorial units for statistics) is a hierarchical system for dividing up the economic territory of the EU. 2 This places France 19th on the OF development rank for European countries.

doubling of area and farms. This could be explained, not only by increased demand, but also by enhanced aid measures for conversion to organic farming, or COF aid, under the Hexagonal Rural Development Programme (PDRH) 2007–2013 (under the European Rural Development Regulation). However, the heterogeneous character of the spatial distribution of organic farms remains substantial (see Fig. 1). Despite a diffusion of OF into new territories and for new productions, there are still OF “deserts”; not only in the cereal plains of the Paris Basin, but also in certain areas located in regions where OF is more prevalent. At the same time, OF continues to grow significantly in the areas where it is already present. Little research has been done on the diffusion and spatial structure of OF in France or even in other countries. Papers by Allaire et al. (2014, 2015) provide an original contribution relating to all the French regions. The first made an exploratory analysis of the diffusion of OF in its spatial and temporal dimension between 1993 and 2009, derived from conversion aid data. Based on various indicators (location quotient – LQ, Gini and Moran indices) from the “ESDA” (Exploratory Spatial Data Analysis) toolbox, different regional and sub-regional spatial dynamics were highlighted. Areas where the development of OF is important could be characterized by clustered groupings of micro-territories with high LQ, or, conversely, the latter could be randomly distributed across the regional territory. The spatial structure of the conversion (with COF aid) thus seems to include phenomena of path dependence in certain contexts. Allaire et al. (2015) studied the dynamics of uneven OF spatial diffusion by analyzing the links between location (regional and local contexts), market access and the propensity of municipalities to accommodate at least one certified organic farm in the year 2010. This paper showed that the network of small towns, the profile of potential consumers, the proximity to certified organic operators downstream, and the diversity of production at the local level were all structural elements of the spatial dynamics of OF, to which spatial and temporal agglomeration effects should be added. To build on this existing work, analyses should be conducted on the causal relationships between territorial factors, the anteriority of OF, and the dynamics of conversion according to production type (Fig. 2). These territorial factors are structural, whether they are related to the natural environment, to the size distribution of farms, to dominant technical and commercial orientations, or (to some extent) to the public zoning policy; and they are institutional (or collective), i.e. related to collective capacities that support individual capacities and are specific (idiosyncratic) to a territory. Based on various databases, the present analysis seeks to examine and explain the role of the latter type of factor in different contexts. 2. Data and methodology The paper uses databases collected by the rural development observatory (ODR):3 • The lists, provided by INAO (French National Institute of Origin and Quality), of certified organic operators active in the third quarter of 2010, issued by certifying bodies responsible for the accreditation of different operators (farmers, processors and distributors), with information about the first year of certification, but not about productive activity. • The lists of beneficiaries of aid for COF, introduced in France in 1993, with the indication of the type of production which is in conversion.

3 ODR prepares and makes available to the public and agreed users information and indicators on agriculture (employment, structures, and types of production), the economy of rural territories, rural policies, environment, and quality signs. It is managed by an INRA unit, to which the authors of this article belong.

Please cite this article in press as: Allaire, G., et al., Territorial analysis of the diffusion of organic farming in France: Between heterogeneity and spatial dependence. Ecol. Indicat. (2015), http://dx.doi.org/10.1016/j.ecolind.2015.03.009

G Model

ARTICLE IN PRESS

ECOIND-2343; No. of Pages 12

G. Allaire et al. / Ecological Indicators xxx (2015) xxx–xxx

3

Fig. 1. Share of organic farms by canton (LAU 1) in France in 2010.

These certified organic operators and the farms of COF beneficiaries are considered at the level of municipalities (considering the location of the head office). This allows the geographical extent and the intensity of contracting on a local scale to be mapped. Given the relative importance of micro-territories without any COF beneficiaries, we employed a hurdle model (Cameron and Trivedi, 2013) that breaks down the analysis into two stages: the first to find the determinants of the presence or absence (i.e. the extent) of COF in micro-territories; and the second to study the intensity of COF measured by the number of contracts within micro-territories with at least one conversion aid beneficiary. In both cases the explanatory variables are intended to represent the location factors and the determinants of

spatiotemporal dependence; in addition, our modelling of the first stage takes neighbourhood effects (“spatial lag”) into account. Following previous results, our aim is to confirm using a sophisticated model that, at the local level, a spatiotemporal dependence exists which makes new OF conversion more likely in places where OF has already developed, and that this applies to all types of production. In addition, we expect variation in the model of diffusion according to the type of production and its particular geography. 2.1. Data used Since the reform of the Common Agricultural Policy (CAP) in 1992 put in place measures to support OF (for

Share of organic area in total agricultural area per type (%)

14 Perfumed, aromac and medicinal plants (PAMP) Fruits

12

10 Grapes 8 Forage areas 6 Fresh vegetables 4 France (general) 2 Arable crops 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

Fig. 2. Evolution of the share of organic agricultural area in total utilizable agricultural area (in %) by type of crops. Source: Agence Bio.

Please cite this article in press as: Allaire, G., et al., Territorial analysis of the diffusion of organic farming in France: Between heterogeneity and spatial dependence. Ecol. Indicat. (2015), http://dx.doi.org/10.1016/j.ecolind.2015.03.009

G Model ECOIND-2343; No. of Pages 12 4

ARTICLE IN PRESS G. Allaire et al. / Ecological Indicators xxx (2015) xxx–xxx

conversion – COF – and for maintaining OF – MOF), organic farming has been regarded as a provider of public goods, particularly at the environmental level. These measures can today be found in all countries of the European Union, with variable application conditions and budgets (Pohl, 2009; Sanders et al., 2011). Other support mechanisms might exist at different territorial levels, but COF aid is the more common scheme of public support. In France, MOF aid is limited to a few regions while COF aid is implemented in the whole national territory. In addition the MOF scheme was only implemented in 2007 and is in competition with tax credit, while the COF scheme has existed since 1994. COF aid is currently paid out over 5 years.4 The data on the COF beneficiaries came from the Agency of Services and Payment (ASP), the paying agency. They are individual (by farm), localized to the municipality and longitudinal (farms have a unique identifier in time unless the holder changes, so one can count these farms only for the year of the first COF aid payment, thereby avoiding duplicates). The amount of aid depends on the surface area involved and on the type of production. This database covers the various programmes implemented in France between 1993 and 2010. In the most recent period, the French outreach programme for OF set up a tax credit for OF industry which was implemented with variable conditions of application over the years. From 2008 onwards, small farms, especially in market gardening production, were found to prefer the tax credit to the COF aid, at least until COF aid increased for this type of production. This reduced the representativeness of COF relative to the OF certified field especially in the case of market gardening whose extent in certain departments may be underestimated. Nevertheless, comparison with other sources (Agence Bio, 2012) indicates that COF aid benefited the vast majority of newly certified organic farms (6642 farms between 2007 and 2010). The different types of COF for the period 2007–2010, the number of beneficiaries and the surfaces in conversion are shown in Table 1. Individual data are difficult to use for economic models when we do not have comparable data on non-organic farms (problems of identification or authorization made it impossible to reconcile our data with other data sources). The geographical completeness of the databases used enables us to analyze the spatial diffusion of OF, by aggregating farms’ information at different territorial levels. Such spatial units under analysis must be small enough that the homogeneity assumption is reasonable, while avoiding too many units void of area in organic conversion, in order to be able to model the intensity of the conversion (the number of contracts). Regarding the present study, the spatial units (micro-territories) were designed at a sub-cantonal geographical scale, by redrawing the cantons (corresponding to NUTS4) according to the affiliation of municipalities in less favoured area (LFA) categories. These are aggregated into three classes: (1) mountainous and high mountain areas; (2) other LFA areas and foothills; and (3) the plains areas (unless this class contained too few farms). These spatial units form “micro-territories” (3699 units for mainland France). Indeed, this affiliation is a significant determinant of the features of production systems, their performances and their insertion in markets. We then considered that the farms in a micro-territory, all things being equal, suffer or enjoy the same constraints or benefits due to their location. They share the same social and institutional context that influences farmers’ strategies. For these reasons we consider

4 COF aid aims to offset the costs related to the period of conversion to organic farming. During the course of this transitional phase, today statutorily fixed at between 2 and 3 years, yield tends to decline while the producer cannot yet benefit from the rents related to the commercialization of organic products. 10 Support revalued in 2009; from 2007 to 2008, the amount was 600 D /ha for market gardening and 350 D /ha for arboriculture.

this scale relevant for studying the spatial structure and dynamic of OF diffusion and the capacities of micro-territories to accommodate farms engaged in OF to a greater or lesser extent. The numbers of unique beneficiaries of COF are observed by micro-territory between 2007 and 2010, for all types of COF and by COF type (according to the four types shown in Table 1 above), thus constituting the explanatory variables. In addition, we model the variable number of simultaneous beneficiaries of COF1 and COF2; these two types of aid are often associated,5 and that combination is characteristic of mixed farming systems. It is clear that the distribution of OF was not uniform across the country during the study period (see Fig. 3). Certain microterritories experienced a high rate of contracting (15–20% of farms), while in others no COF measures had been initiated (42% of microterritories). 2.2. Presentation of models As already mentioned, we employed a hurdle model that breaks down the analysis into two stages. The first stage corresponds to a binary process and is modelled by a spatial probit to find the determinants of the presence of COF in a micro-territory. This presence or absence is considered to be a “barrier” (hurdle) in the process, fixing its extent. Once this barrier is crossed, we observe positive counts (number of COF). This is modelled using a zero-truncated negative binomial regression (ZTNB). The first stage, or Model I, refers to the extent of contracting a COF, while the second stage, or Model II, corresponds to the intensity of the phenomenon. The likelihood of the hurdle model is written as follows: f (Y = k) =

⎧ ⎨ f1 (0)

if k = 0

⎩ (1 − f1 (0)) f2 (k)

1 − f2 (0)

if k = 1, 2, . . .

where f1 (0) represents the probability of a zero in the binary part of the model (the spatial probit) and Y a stochastic variable which f2 (k) takes count values. f2 is the likelihood of a 1−f particular count (0) 2

(a negative binomial model), which makes the probability distribution for the truncated count (by the negative binomial model truncated at zero). The factor (1 − f1 (0)) in the second equation is there to assure that the sum of probabilities amounts to one. The two parts of the model are independent, so the maximum likelihood estimation of the hurdle model can be done by separately maximizing the two likelihood terms: one for the model of zeros and one for the model with positive counts. The spatial probit of the first stage enables one not only to find the determinants of whether or not COF is present in microterritories, but also to study the spatial dependence of these territories relative to their neighbours. The spatial dependence is introduced in an autoregressive manner (Chapter 10 in Lesage and Pace, 2009), and the model is: y = Wy + Xˇ + ε,

with

ε∼N(0, ε2 In )

where y is the binary variable indicating the presence of farmers with COF contracts, and W is a neighbourhood matrix. The specification used for W is a row standardized contiguity matrix. Thus, two units are neighbouring if they are adjacent or contiguous, and the weights of the matrix W are standardized so that each row sums to one. With individual data, this type of specification can explain the choice of involvement or not of farmers in the COF territory as a weighted linear function of the average of the decisions of farmers’

5 85% of the beneficiaries of COF1 also engaged with COF2 surfaces; the inverse is true for 64% of recipients of COF2.

Please cite this article in press as: Allaire, G., et al., Territorial analysis of the diffusion of organic farming in France: Between heterogeneity and spatial dependence. Ecol. Indicat. (2015), http://dx.doi.org/10.1016/j.ecolind.2015.03.009

G Model ECOIND-2343; No. of Pages 12

ARTICLE IN PRESS G. Allaire et al. / Ecological Indicators xxx (2015) xxx–xxx

5

Table 1 Amount, beneficiaries and areas concerned by a conversion to organic farming aid (COF) for the period 2007–2010 in relation to the type of crops. Type of COF (depending of the type of crops)

Amount (D /haa )

Total no. of COF beneficiaries

COF1 COF2 COF3 COF4 All COF

100 200 350 900

2556 4153 1871 1208 6642

a

Permanent grassland (and chestnut groves) Annual crops and temporary grassland Field vegetable crops, viticulture and perfumed, aromatic and medicinal plants (PAMP) Market gardening and arboriculture10

Area concerned (ha) 74,561 133,964 19,686 5699 233,910

D : euro; ha: hectare.

Fig. 3. Number of COF aid beneficiaries by microterritories from 2007 to 2010 according to the four types of COF.

Please cite this article in press as: Allaire, G., et al., Territorial analysis of the diffusion of organic farming in France: Between heterogeneity and spatial dependence. Ecol. Indicat. (2015), http://dx.doi.org/10.1016/j.ecolind.2015.03.009

G Model ECOIND-2343; No. of Pages 12 6

ARTICLE IN PRESS G. Allaire et al. / Ecological Indicators xxx (2015) xxx–xxx

neighbours Wy (spillover effect) and of individual explanatory variables X. In this case, when one thinks about the characteristics of the territories and not just of the characteristics of the farms within them, the neighbourhood effect, or spatial dependence, covers both spillover effects (of individual decisions) expressed at the level of aggregated micro-territories and the effect of micro-territorial institutions that impact on neighbouring territories. The formulae of the likelihood of the truncated negative model are specified by Hilbe (2011), Chapter 11. The choice of this model is related to the fact that the rates of contracting are low, the majority of micro-territories (73%) that contain a farm benefiting from a COF have no more than three contracting farms. Location factors are integrated in the models by the characterization of micro-territories in terms of natural and social resources (governance and institutional capacity) and production systems (including marketing methods and strategies). The dependence effects are identified by variables related to the history of organic farming in the territory: the number of years organic farms have been present; and the presence of certified organic downstream operators (processors, wholesalers and retailers). We focused this analysis on institutional factors, while controlling for economic factors involved in the geographical diffusion of OF. Apart from the zoning of LFAs that reflects natural factors, we constructed two sets of explanatory variables designed to capture: • the strategic orientation of production systems: dominant production, level of inter-farm diversity, importance of signs of quality other than OF; • the institutional factors: territory belonging to zoning regulation (Natura 2000, etc.), professional organic organization, OF leadership and support structures at the local level, principal measures of the second pillar contracted during the period (2000–2006) prior to the observation period. Several limitations constrained the construction of variables. These related to the databases used and to the choice of observation level (the aggregation of certain variables limits their explanatory power due to micro-territorial heterogeneity). But by using several sources and by performing the appropriate tests, the proposed model is consistent from a theoretical, empirical and technical point of view. As well as the databases from COF and certified operators, several other sources were utilized: the database of all the aid of the second pillar of the CAP under the previous national rural development programme (PDRN) (2000–2006) (source ASP/ODR)6 ; the database of the farmers subjected to the mandatory social security scheme (base termed MSA/ODR, in this paper), as well as various statistical and agricultural sources (including the agricultural Census – RA2000). The models presented are modified according to the type of COF (see Table 1), for all COF and for the combination of COF1 and COF2 on the same farm. This results in a total of six independent estimations of each model. 2.3. Variables used in Model I (extent) and Model II (intensity) Explanatory variables statistics are given in Table 2 (see also the maps Fig. 3). In this table, the number and rate of concerned microterritories refer to the first stage of the study (Model I – extent), while the other lines give information about the distribution of the explanatory variable of Model II (intensity). A description of all variables used in the different models, including their abbreviated names, is available in Appendix. The

6 ASP is the French paying agency for the second CAP pillar. The data were collected by ODR (see footnote 3).

variable on agricultural orientation of micro-territories (“OTE”) was constructed based on the declarations made by farmers on the type of principal activity for compulsory work accident insurance (source MSA/ODR) in 2007. At the micro-territorial level, an OTE is considered dominant if it represents at least 60% of the area of farms and half of farmers (the territory is therefore specialized). A variable in six classes was constructed (crops, fruits and vegetables, viticulture, dairy cattle and mixed cattle, beef cattle, and areas without a dominant orientation). The variable on the number of farms (LOG FARM07) allows us to control for the size effect of a micro-territory. Diversification strategies are expressed in three variables which concern the involvement of farms in direct sales, farm processing, and tourism/catering activities, according to RA2000. Another variable indicates the impact of CUMA, which are local cooperatives for the utilization of agricultural material. The importance of official signs of quality and of origin is taken into account via five variables, which refer to the presence (extent model) or to the number (intensity model) of farms authorized for: Product Conformity Certification (PCC); Label Rouge (Red Label – RL); wine Protected Designation of Origin (PDO); non-wine PDO; and other signs. These data are also from the RA2000 and thus characterize past strategies (thereby avoiding an endogeneity bias). The variable referring to the presence of organic certified processors or distributors in 2006 was elaborated thanks to data from the INAO. Environmental issues in the territory, as reflected in the French Rural Development Programme, are represented by three variables. Two involve water: the presence of a “Grenelle catchment”7 and the presence of public measures under the Water Framework Directive (in French: DCE, Directive Cadre Eau). A third captures biodiversity, through the areas related to Natura 2000 sites. To account for the heterogeneity of public policies and the capacities of territories to take the opportunities they offer, two types of proxy were used. First, the relative concentration of organic farms on a larger level in a micro-territory, calculated using the location quotient (LQ) of the department in relation to France as a whole. This specifically indicates differences in outreach programmes for OF community development conducted by the departmental chambers of agriculture and specific nongovernmental organizations. Second, the type of rural development measures contracted under the previous rural development programme (PDRN) within a micro-territory for the 2000–2006 period. The nature and intensity of past contracting reflects the nature of local developmental and environmental issues, but also the collective capacity within the territory to seize the opportunities offered by public policies. The 22 measures of PDRN were regrouped, to avoid problems of correlation between some of them and to ensure analytical consistency; seven types are thereby distinguished as binary variables (indic typoX 1) or as variables indicating the number of total beneficiaries (rdp1farm typoX), according to the stage of modelling. Following the same numbering as the corresponding variables, the seven types are: (1) The entry-level measures (areas with handicaps and extensive livestock farming), oriented towards income support. (2) The agri-environmental measures concerning water, soil and biodiversity issues. (3) Aid related to forestry. (4) Aid for investment in farms and the setting-up of young farmers.

7 507 priority catchment areas (also called “Grenelle catchments”) were defined in 2007 because of serious threats of diffuse pollution, especially nitrates and pesticides. For each of these catchments a sensitivity area was defined, for which an action plan was developed and implemented in 2012.

Please cite this article in press as: Allaire, G., et al., Territorial analysis of the diffusion of organic farming in France: Between heterogeneity and spatial dependence. Ecol. Indicat. (2015), http://dx.doi.org/10.1016/j.ecolind.2015.03.009

G Model ECOIND-2343; No. of Pages 12

ARTICLE IN PRESS G. Allaire et al. / Ecological Indicators xxx (2015) xxx–xxx

7

Table 2 Number of micro-territories with one or more COF aid beneficiaries between 2007 and 2010.

Number of micro-territories concerned Rate of micro-territories concerned (total = 3699) With one COF beneficiary (% of micro-territories concerned) With two COF beneficiaries (% of micro-territories concerned) With three or more COF beneficiaries (% of micro-territories concerned)

All COF

COF1

COF2

2162 58% 825 (38%) 457 (21%) 880 (41%)

1282 35% 682 (53%) 293 (23%) 307 (24%)

1712 46% 774 (45%) 400 (23%) 538 (31%)

(5) Measures concerning the food industry (notably cooperatives). (6) Aid for training, facilitating, counselling and support for rural development. (7) Social and rural measures, in particular aid for early retirement, farm diversification and services in rural areas. The issue of spatial and temporal dependence is treated at the level of micro-territories by a variable that indicates the presence before 2007 in the micro-territory of at least one operating organic farm (variable constructed from INAO data on certified organic operators and from ASP data on COF aid between 1993 and 2006). The spatial autocorrelation parameter rho expresses the neighbourhood spatial dependence; it represents the simultaneous effect of the value taken by the dependent variable in the micro-territory and in the adjacent micro-territories. 3. Results and discussion The first model accounts for the spatial extent of OF (presence or absence of COF contracting between 2007 and 2010; see Table 3). The second reflects the intensity of contracting in the areas concerned (see Table 4). In both models a control variable takes into consideration the importance of agricultural activity in micro-territories [LOG FARM07]. 3.1. Spatial structure of OF: between expansion and concentration The length of time OF has been present in micro-territories [timeOF07bin] before the studied period has a positive impact on their propensity during the period 2007–2010 to accommodate at least one new beneficiary of COF (Model I), regardless of the type. In Model II, we again find this effect on the intensity of contracting. Adding this to the departmental effect [LQdepfce sup006] (a high location quotient at departmental level has a positive effect locally on the number of beneficiaries), we see that the process of OF concentration depends on both local years’ experience in OF farming and departmental support. However, the temporal dependence is less pronounced when considering the intensity of COF3 and COF4 contracting. This is due to the fact that a large share of conversion for vine (COF3) and, moreover, for arboriculture (COF4) are recent (see, in Fig. 1, the strong growth in OF areas for vineyards and orchards during the studied period). For COF4, the recent development of an organic fruit and vegetable sector and the 2008 reassessment of the amount of COF4 aid may explain the low local temporal dependence and thus a movement of spatial expansion; however, only within territories specialized in the production systems concerned. The parameter rho, which expresses the level of correlation between neighbouring territories, is significant for all Models I, showing a clustering trend characterizing the studied period (already indicated for previous periods in Allaire et al. (2014)). These clusters are sub-departmental but include several microterritories. This spatial dependence parameter has a stronger value for COF2 and COF4, and a lower value for COF1 and COF3. The COF2

COF3 695 19% 386 (56%) 101 (15%) 208 (30%)

COF4 663 17% 450 (68%) 107 (16%) 106 (16%)

COF1 * COF2 1100 30% 605 (55%) 257 (23%) 238 (22%)

type includes both annual crops (but is not well represented in specialized cereal areas such as the Paris Basin) and temporary grassland, which is well represented in the dairy areas of Brittany and Pays-de-la-Loire (see Fig. 3). As evidenced by the model for the COF1/COF2 combination, these clusters also correspond to areas of mixed crop/husbandry systems. For COF4, if there is no temporal dependence, there is in contrast an apparent tendency to spatial concentration related to the specialization of fruit production areas (of which the size exceeds our micro-territories). The opportunity for conversion is dependent on the characteristics of production basins, related to the capacity of local cooperatives and other intermediaries to support organic production and to the availability of technical support networks, fruit organic production being exposed to significant technical risks. In the case of COF3, the clustering phenomenon is significant but weaker, while OF diffusion occurs in all wine-making territories. The effect of the presence in the micro-territory of downstream organic operators [dwstrof20061] does not appear clearly for Model I, except for COF 3 with a positive effect. Since this variable includes all kinds of processors and others certified downstream operators, it is difficult to interpret as a factor of geographical OF extension. But it impacts on the intensity of COF contracting in the Model II for all the COF types, showing the existence of organic districts. For a micro-territory, being located in a simple LFA or in the foothills [fa3c12] has a positive impact on the intensity of COF contracting in general, and in particular for COF2 and COF4. In contrast, being located in mountainous or high mountain areas [fa3c34] has no overall effect, but a negative effect for COF1. Several factors could explain this situation: a lower attraction for OF in mountainous areas (more difficult to access markets, etc.), but also possible antagonism between different development schemes. Statistical relationships established by these first results show a geographical dependence of the spatial structure of the OF areas established during the period 2007–2010 related both to the geographical constraints and to the type of production, particular geographies and phenomena of clustering which reflect and confirm the role of the social context and experience in the diffusion of OF.

3.2. Productive specialization of territories, market strategy and intensity of COF contracting In Model I, the relationship between the agricultural orientation of a micro-territory and the presence of COF was determined by a variable related to the dominant orientation (note that 48% of micro-territories have no dominant specialization). The results show that the type of COF tends to conform to the territorial specialization, if it exists. Thus, specialization in viticulture [OTE3738] is specifically favourable for COF3 and unfavourable for COF1 and COF2. Similarly, COF4 contracting is greater in areas dominated by fruits and vegetables [OTE23]. This effect also influences the intensity of contracting. This result is coherent with the previous result regarding these productions. Micro-territories where “annual crops” is the dominant activity [OTE1] are not generally favourable to the presence of beneficiaries of COF and significantly not for COF1 (permanent grassland).

Please cite this article in press as: Allaire, G., et al., Territorial analysis of the diffusion of organic farming in France: Between heterogeneity and spatial dependence. Ecol. Indicat. (2015), http://dx.doi.org/10.1016/j.ecolind.2015.03.009

G Model

ARTICLE IN PRESS

ECOIND-2343; No. of Pages 12

G. Allaire et al. / Ecological Indicators xxx (2015) xxx–xxx

8 Table 3 Results for model I on the extension of COF aid uptake. VARIABLE (d if dummy)

Toute COF

INTERCEPT LOG FARM07 OTE1 (d) OTE23 (d) OTE3738 (d) OTE43A (d) OTE4B (d) timeOF07bin (d) fa pccf1 (d) fa rlf1 (d) fa othf1 (d) fa pdo nowinf1 (d) fa pdo winf1 (d) pres catch grenl1 (d) Natura2000 (d) eliDCE (d) indic typo1 1 (d) rdp1farm typo2 indic typo3 1 (d) rdp1farm typo4 indic typo5 1 (d) indic typo6 1 (d) indic typo7 1 (d) dwstrof20061 (d) RHO BPT BP1 N AUC Rate

−0.8025 0.5203 −0.0824 0.0844 0.1789 −0.0033 −0.479 0.6252 0.1417 −0.0797 0.05 −0.0032 0.2205 0.0455 −0.0487 −0.0507 0.1969 −0.1749 0.0695 0.4112 0.0512 0.1498 0.0945 0.0252 0.1628

COF 1 *** ***

*** *** **

***

**

**

*** 0.738 0.773 3699 0.8 0.584

−1.8673 0.3504 −0.3206 −0.6357 −0.6802 0.322 −0.1773 0.435 0.1056 −0.0426 −0.0103 0.0857 −0.1044 0.1383 −0.0504 −0.1671 0.5764 −0.1039 0.1942 1.2375 −0.1235 0.1749 0.2638 0.0111 0.1688

COF 2 *** *** *** *** *** – *** –

– – – * *** ** *** * *** *** *** 0.715 0.8 3699 0.81 0.347

−1.1407 0.487 0.1135 −0.6148 −0.5153 −0.1474 −0.5676 0.5003 0.0815 0.0474 0.0392 −0.0194 0.0115 0.0841 −0.095 0.0736 0.1681 −0.0285 0.0788 1.2094 −0.005 0.0408 0.2173 0.0787 0.243

COF2, which includes annual crops and temporary grasslands, corresponds more to livestock farms or mixed farming than to specialized grain farms and consequently is not correlated to annual crops specialization areas. At the individual level, two-thirds of COF2 are associated with a COF1 and, inversely, 83% of COF1 (permanent grassland) is associated with a COF2 (temporary grassland

COF 3 *** ***

*** * *** ***

* *

***

*** *** 0.692 0.754 3699 0.77 0.463

−1.5339 0.4242 −0.1268 −0.0629 0.7508 −0.4224 −0.5954 0.513 −0.0313 −0.1352 0.0495 0.0514 0.5579 0.0525 0.1549 −0.0661 0.1285 −0.3869 −0.2532 0.1054 0.1904 0.1127 0.03 0.1393 0.1219

COF 4 *** ***

*** *** *** *** –

*** **

** ** * * * 0.748 0.727 3699 0.82 0.188

−1.2869 0.3301 −0.1669 0.6502 0.1276 −0.1343 −0.3589 0.3334 0.1005 −0.0895 0.3866 0.2487 0.0786 0.0116 −0.0945 −0.1242 0.0728 0.361 −0.073 −1.5511 0.2088 0.0941 0.0965 0.0547 0.2614

COF 1 and 2 *** *** * * – ** ***

*** ***

–

*** ***

*** 0.665 0.733 3699 0.75 0.179

−1.9711 0.3829 −0.3204 −0.3423 −0.8455 0.126 −0.3151 0.371 0.0962 0.0026 0.083 0.0645 −0.1474 0.1982 −0.1232 −0.0647 0.5963 −0.2061 0.1643 1.5128 −0.1453 0.111 0.3068 0.0423 0.2195

*** *** *** *** – ** *** –

* * * *** * *** ** * *** *** 0.699 0.797 3699 0.8 0.297

and self-production of organic grain for the supply of feed for organic herds). Micro-territories specializing in dairy and mixed cattle [OTE43a] have a higher propensity for COF1, but lower for COF2; in contrast, they have a higher probability of having individuals simultaneously contracting for COF1 and COF2. Specialization in meat cattle [OTE4b] reduces the propensity to accommodate

Table 4 Results for model II on the intensity of COF aid uptake. VARIABLE (d if dummy)

All COF

INTERCEPT LOG FARM07 lfa3c12 (d) lfa3c34 (d) MONO071 (d) shar eliDCE shar natura2000 nbfa pcc nbfa rl nbfa oth nbfa pdo nowin nbfa pdo win shar dirsal shar process shar wcuma Tourcater timeOF07 dwstrof20061 (d) rdp1farm typo1 rdp1farm typo2 rdp1farm typo3 rdp1farm typo4 indic typo5 1 (d) rdp1farm typo6 rdp1farm typo7 LQdepfce sup006 shar ate12 07 AIC BIC N

−0.903 0.81 0.315 0.026 −0.017 −1.175 −0.047 0.947 −1.036 0.29 0.234 0.609 0.29 −0.225 0.359 0.051 0.052 0.114 −0.257 0.49 −0.828 1.019 0.189 2.726 −0.248 0.189

COF 1 *** *** ***

*** – ** – ***

*** * * * ** *** ** *** 7552.1 7705.4 2162

−1.392 0.784 −0.085 −0.426 −0.108 −0.642 −0.514 0.795 −1.82 −1.412 −0.144 −0.799 0.429 −0.383 1.2 −0.081 0.055 0.078 0.618 0.705 0.066 0.673 −0.178 3.89 4.989 0.157

COF 2 *** *** * – * *** ** –

*** *** * *

* ** * *** 3311.7 3450.9 1282

−1.378 0.736 0.426 0.224 0.035 −0.773 −0.475 0.654 −1.691 −0.635 −0.29 −1.136 0.484 0.808 0.963 0.01 0.059 −0.007 −0.347 0.582 −0.774 1.711 0.033 1.258 2.181 0.276

COF 3 *** *** ***

** * *** * *** * *** *** – * * ***

*** 5155 5302 1712

−1.516 0.726 0.03 0.025 0.023 −2.106 0.061 3.078 −5.899 0.635 0.995 0.91 1.195 0.143 −1.94 0.211 0.022 0.073 −0.919 −0.548 −2.774 2.938 0.73 5.014 −11.74 0.111

COF 4

COF 1 and 2

*** ***

−2.07 0.863 0.74 0.249

*** *** ***

**

−2.871 −0.178 2.711 2.256 −0.512 1.532 1.037 −0.324 −1.768 −1.59 0.646 0.016 −0.13 −0.601 −0.34 −4.392 −0.386 0.139 7.79 0.05 0.071 3.346

**

* ** * *** – * –

– – ** *** * * 2023.8 2146.5 695

* – ** * – – ***

*

**

*** 1398.2 1519.6 663

−1.689 0.765 0.066 −0.171 0.005 −0.626 −0.593 0.288 −1.696 −1.872 −0.741 −1.286 0.613 0.552 1.161 −0.085 0.052 0.074 0.372 0.763 −0.936 1.145 −0.18 3.863 8.096 0.197

*** ***

– * ** ** * *

*** ***

* * * ** *** *** 2705.4 2840.5 1100

Please cite this article in press as: Allaire, G., et al., Territorial analysis of the diffusion of organic farming in France: Between heterogeneity and spatial dependence. Ecol. Indicat. (2015), http://dx.doi.org/10.1016/j.ecolind.2015.03.009

G Model ECOIND-2343; No. of Pages 12

ARTICLE IN PRESS G. Allaire et al. / Ecological Indicators xxx (2015) xxx–xxx

beneficiaries of COF in general, and of COF1 and COF2 in particular (alone or combined). Milk production is a better structured industry from a collection and processing perspective, and is more favourable than the meat industry to the diffusion of OF. Moreover, in the years preceding the period studied (2007–2010), a saturation of the local markets for meat (the main way in this case to enhance the choice for OF) occurred as a result of an “anarchic” wave of conversion to organic farming in the wake of the BSE crisis; the reverse is true for the organic dairy industry, which, after a period of stagnation, has recently experienced significant growth (A.N.D.-I., 2008). The effect of a mono-oriented or non-local production system [MONO071] (criterion indicating a high degree of specialization) is not apparent from Model II because only 38% of micro-territories are concerned. Moreover, when one would expect a specialization effect, e.g., for wine production and COF3, this effect is doubtless captured by another variable, that is, the number of farms engaged in wine PDO [nbfa pdo win]. The number of farms involved in a quality scheme in 2000, PDO [nbfa pdo nowin; nbfa pdo win] or other labels category (including especially the PGI) [nbfa pcc; nbfa rl; nbfa oth], subsequently has a negative effect on the intensity of joint COF1 and COF2 contracting, and on each of them individually for the variable “other labels” [nbfa oth]. These effects relate to the nature of the production systems involved in these types of COF, which do not correspond with the use of market quality signs. However, there is no general antagonism between OF and Geographical Indications (GIs: PDO or PGI), because there is no effect demonstrated in Model I which could show that territories with GIs are averse to contracting a COF. Regarding wine PDOs [nbfa pdo win] (which involve more than 57% of the areas with vines), there even exists a positive effect on the extent of COF3: OF appears, in the latest period, as a compatible and even complementary strategy to PDO, in a context where wine growers seek competitive advantages through differentiation. Labels Rouges (Red Label), which mainly concerns livestock raised for meat, [nbfa rl; fa rlf1], have no effect on the extent and a negative impact on the intensity, for all types except COF4 (note that the label applies to very few fruits or vege). It seems that there is an incompatibility between the two strategies (OF and Label Rouge); incompatibility of specifications in certain cases and, without doubt, of image too. This arises since there is a risk that the differentiation between organic and non-organic Label Rouge products will eventually tarnish the image of the label or will not allow further value-creation for organic labelled products. The CCP (certificate of compliance, private label) [nbfa pcc; fa pccf1] approach, at the level of micro-territories, has a positive effect in Model I on the COF extent in general (COF1 and COF1 and COF2 combined -COF1*2- especially) and in Model II on the intensity (of COF4 in particular). This effect is, however, difficult to explain because it involves many products and micro-territories. It may indicate a capacity for collective action. The spatial structure of OF, which differs according to geography and the types of production, is impacted at the local level by the degree and nature of farming specialization and by the existence of opportunities to differentiate the products by characteristics which are either alternative or complementary to OF. 3.3. Identification of the effects of collective capacities and strategies The previous analysis of the effects of the market quality signs concerns variables calculated from RA2000 and aggregated at the micro-territorial level, thus characterizing specific agricultural systems that existed before the studied period (2007–2010) and which certainly had some continuity. They express a collective experience of a long duration, which has shaped collective capacities.

9

To complete the characterization of the idiosyncratic capacities of micro-territories, other aggregated variables on diversification strategies were introduced, also derived from RA2000: the share of farms involved in direct sales [shar dirsal] has no effect (low significance) on the contracting of COF3, in keeping therefore with the strategies of wine growers and rural development policies promoting direct sales (“wine routes”, for example). The importance of processing activities on the farm [shar process] applies positively to the contracting of COF2 (these include dairy areas in particular), and negatively to COF4 (fruits and vegetables). Tourism and catering activities on the farm [Tourcater], more prevalent in the southeast, have a positive effect on the numbers of COF4 and COF3 contracts. These results confirm the effects related to the orientation of the local production systems previously highlighted, showing that the territories formerly committed to the valorization of agricultural production by quality schemes or para-agricultural activities are likely, even in recent times, to invest more easily in OF, which is itself a strategy of value creation. One variable, also calculated in 2000, refers more directly to the local cooperation practices between farmers [shar wCuma]. It is the importance of work provided by CUMA (cooperatives for the use of agricultural equipment) which has a positive effect for COF1 and COF2 on the number of beneficiaries, and negative for COF3 and COF4. This corresponds to the fact that the use of CUMA is more common in livestock farms in the west than in the agricultural systems of the southeast. Variables constructed from the aid granted under the preceding programme of Rural Development Regulation (PDRN 2000–2006), regrouped into seven categories, indicate that territories do not have the same ability or opportunity to seize public policies. Measures related to the compensation of a natural handicap (ICHN) or to the extensive nature of livestock production (PHAE) [indic typo1 1; rdp1farm typo1] increase the propensity of having beneficiaries of a COF, and their number, for COF in general and more specifically for COF1 and COF2. Here, once again, we find the geographical effect mentioned for foothills and LFAs which are those corresponding to these two measures, apart from mountain areas. The link with investment measures [indic typo4 1; rdp1farm typo4] is more complex: it is negative for the propensity to accommodate at least one recipient, but it is positive for the intensity. Related to specific investments at the downstream level, the aid measure for food industries (POA measure) [indic typo5 1] has a strong positive effect for COF3. This can be explained by the fact that cooperative wineries frequently use this aid for investment in developing structures for receiving organic production. Finally, the support related measures (training, etc.) [indic typo6 1; rdp1farm typo6] have a positive effect on the extent and intensity for the COF in general. The set of measures that concern rural but not agricultural production [indic typo7 1; rdp1farm typo7] has a positive effect for COF1 and COF2 (reflecting local economic dynamism) and negative for COF3 (vine areas have benefited little from these measures). Agri-environmental measures (AEM), apart from mass measures such as extensive grazing support, respond to issues related to water, soil or biodiversity and include both the COF and, more frequently, measures concerning fertilization or pesticides. They were not focused on designated areas for the period 2000–2006. The increase in their level of contracting [indic typo2 1; rdp1farm typo2] raises the propensity of micro-territories to accommodate more beneficiaries of a COF (Model II), and more particularly a COF1 and COF2. This effect is apparently contrary to the effects of agrienvironmental zoning priorities introduced from 2007 with the implementation of the second European Rural Development Regulation (PDRH). From then, these “territorialized” AEM could only be contracted in priority areas, called TAEM areas, covered by a local agri-environmental project. In these areas, the conversion to organic farming is encouraged and various measures, less

Please cite this article in press as: Allaire, G., et al., Territorial analysis of the diffusion of organic farming in France: Between heterogeneity and spatial dependence. Ecol. Indicat. (2015), http://dx.doi.org/10.1016/j.ecolind.2015.03.009

G Model ECOIND-2343; No. of Pages 12 10

ARTICLE IN PRESS G. Allaire et al. / Ecological Indicators xxx (2015) xxx–xxx

demanding, are proposed. It appears that the greater the share of a micro-territory surface classified as TAEM area (with environmental issue “water”) [shar eliDCE], the less intense contracting (Model II), for all types of COF. Several explanations can be offered for this apparent contradiction. First, the two periods of contracting do not cover the same territories;8 indeed the PDRN (2000–2006) has been criticized for the lack of targeting AEM in relation to the importance of the issues. We can say that the AEM of PDRN were more frequently contracted where, for various reasons, there was an environmental sensitivity (thereby explaining the path dependence) than where there were major environmental problems. Thus, there have been few contracts for these measures in Brittany, where today one finds numerous TAEM areas with water as an environmental issue. Second, these areas are often located in areas of major crops and intensive breeding, with production systems that are less inclined than others to OF conversion. The presence of a Grenelle catchment area in micro-territories [pres catch grenl1] has no effect, or a depressive effect. This is logical in the sense that these catchments were identified in 2009 as high-risk areas in terms of pollution (particularly related to intensive agricultural practices such as use of nitrates and pesticides). Second, specifically when it comes to areas without experience of AEM and OF, the transformation of local production systems cannot be immediate but takes time. In addition, the mid-term review (2010) of agri-environmental projects is generally not considered satisfactory, 7–8% of the eligible area was engaged in an AEM with water as the issue (source ODR). The situation is somewhat different in TAEM areas with Natura 2000 as an issue [Natura2000; shar natura2000], which are not inherently related to intensive agriculture, and have long aimed at local cooperation and experience a better overall AEM contracting rate. One also notes in this case the absence of links with the presence of COF beneficiaries; when an effect exists, it is also negative. The political will to promote OF in priority areas has not been translated into statistical facts, at least for the period up to 2010. However, we must be cautious in assessing local AgriEnvironmental projects because, on the one hand, local conditions are highly variable and, on the other, the level of observation used here is not the most appropriate for deriving conclusions on the effectiveness of targeted environmental programmes. This would require comparisons of the evolution of territories with the same characteristics, regardless of whether or not they are located in a TAEM area. If these observations indicate that the zoning of environmental priorities alone does not produce effects, they do not invalidate, and may even support, the existence of phenomena of spatiotemporal dependence in the diffusion of OF in France. To reveal the role of territory-based collective capacities we used, on the one hand, variables calculated from RA2000 and aggregated at micro-territorial for characterizing agricultural socioeconomic systems existing before the studied period (2007–2010) and, on the other, the responses to previous rural development policies. Both these were proved to reflect local social dynamics. 4. Conclusion By analyzing the extent and the intensity of contracting COF aid in France at the micro-territorial level, and by type of production, the model presented has illustrated the variability of the mechanisms involved in the territorial diffusion of OF between 2007 and 2010. The results show the importance of considering, beyond individual behaviour, the collective capacity of territories for OF conversion. First, we must emphasize the importance of

8 In the case where it is the same territories, taking into account that the duration of contracts is 5 years, the importance of contracts signed from 2003 up to 2006 reduces the possibility of contracting for 2007–2010.

spatial dependence, or, more precisely, spatiotemporal dependence, which remains significant, despite diffusion into new territories and new specializations (such as market gardening and arboriculture): the territories where OF is already present have the highest propensity to conversions, with a higher intensity. The presence of certified organic downstream operators reinforces this. The contracting of COF is also more likely in neighbouring microterritories, confirming a clustering tendency at this level, stronger or weaker depending on the type of production. It is legitimate to speak of path dependence, which shows here that experience at the territorial level promotes the engagement of farmers in OF. Spatial heterogeneity is found in a combination of factors, which marks the specialization (or indeed the non-specialization) of territories; specialization of production systems (wine region, fruit and vegetable area, or dairy area), but also specialization in farms’ development strategies (greater or lesser importance of different quality signs, farm processing and direct sales, etc.). The agriculture of micro-territories that promote COF already seems more oriented towards end consumers than that of others; it is the existence of capacities corresponding to this orientation that explains the spatiotemporal dependence. The analysis of the influence of past institutional experience shows that territories developed different skills and strategies in a collective manner which marked not only a heterogeneity among regions, but also led them down a path that influenced their future choices and capabilities. These findings on spatial dependence are consistent with the forecast established in a note by the Centre for Studies and Prospective of the Ministry of Agriculture (Mahé, 2012) on the prospects of OF by 2015 (based on individuals’ intentions to convert, collected during the agricultural census of 2010). It specified the characteristics of farms that intended to convert to organic farming. These are similar to those of farms already engaged in OF: profile of the farmer (younger, higher level of education); type of production (higher proportion of herbivorous livestock for example); extraagricultural activities (more highly engaged in processing, rural tourism, and direct sales); and location (intentions to convert occur more frequently in areas already relatively high in OF, especially in the southeast). However, if public authorities want to promote OF to meet the high targets set, they must establish a coherent strategy – in particular to develop collective capacities, specific to each territory. In reality, attaining 20% of the UAA in OF (or even 10%) will only be possible if organic practices spread into new territories, where they remain very insignificant. Nevertheless, the financial strategy (COF-type aid), given that it is generally effective9 , is still not sufficient to meet all the challenges and to take into account the specificities of each territory and its production system. Some territories must respond to strong local environmental issues (in addition to global issues such as climate change). These might be the protection of water resources, biodiversity, etc. Furthermore, productive and strategic orientations should be considered under the potential for conversion. OF must be a credible technical and commercial opportunity for local production systems. This involves the development of the sector (markets), remunerative prices, consistency and complementarity with other development strategies, and the implementation of an overall strategy to integrate OF and the various public measures in relation to environmental issues. Based on this study, for this reason, we point out the necessity of developing specific support (not only financial support but also technical and commercial counselling and training; incentives for value chain development, etc.) focused on micro-territories where OF is still under-developed. Indeed, to activate the path dependence

9 The general effectiveness is shown by the renewed conversion following the re-evaluation of aid, particularly for COF4.

Please cite this article in press as: Allaire, G., et al., Territorial analysis of the diffusion of organic farming in France: Between heterogeneity and spatial dependence. Ecol. Indicat. (2015), http://dx.doi.org/10.1016/j.ecolind.2015.03.009

G Model

ARTICLE IN PRESS

ECOIND-2343; No. of Pages 12

G. Allaire et al. / Ecological Indicators xxx (2015) xxx–xxx

in favour of OF extension, it is necessary to help territories, each of which has its own specificities (in terms of production and markets), to enter in this dynamic. In addition, for those territories where OF is already present, policy supports should be oriented to counter any barriers, due, for example, to an insufficient development of logistic and marketing solutions, or to the antagonism with other quality labels. However it could be difficult to define a general strategy, since the development of OF results from a complex interaction among various factors (technical, socio-economical and institutional) that differ from place to place.

Description

Number of farms in 2007 (log(var)) OTE1: Dummy indicating that ‘field-crop’ type of farming is dominant OTE23: Dummy indicating that ‘fruits and vegetables’ type of farming is dominant OTE3738: Dummy indicating that ‘viticulture’ type of farming is dominant OTE43a: Dummy indicating that ‘dairy cattle’ or ‘mixed cattle’ type of farming is dominant OTE4b: Dummy indicating that ‘beef cattle’ type of farming is dominant Dummy variable indicating that one crop covers more than 50% of the farms and more than 60% of the area in the micro-territories Share of farms with market gardening or arboriculture in 2007 (log(var + 1))

Share of farms with direct sales Share of farms with on-farm processing Share of farms using labour from CUMA (Cooperatives for the use of agricultural equipment)

Acknowledgments The authors acknowledge the support of FP7 collaborative project SPARD (Spatial Analysis of Rural Development Measures). The usual disclaimer applies.

Appendix. Table 5: presentation and descriptive statistics for variables of interest in Models I and II

Variable for the spatial probit model (Model I – extension)

Variable for the negative binomial truncated model (Model II – intensity)

Descriptive statistics for the complete sample (unlog-transformed and uncentred variables)

Name

Type

Name

Type

Dummy variable (share of positive)

LOG FARM07 OTE1

Continuous Dummy

LOG FARM07

Continuous

OTE23

Source

Continue variable average (standarddeviation) 122.5 (94.6)

16%

MSA MSA/ODR

Dummy

7%

MSA/ODR

OTE3738

Dummy

6%

MSA/ODR

OTE43a

Dummy

24%

MSA/ODR

OTE4b

Dummy

6%

MSA/ODR

38%

MSA/ODR

Less favoured areas (LFA): simple and foothills LFA LFA: mountain and high mountain pres catch grenl1 Dummy Presence of at least one priority “Grenelle” water catchment eliDCE Dummy Share or presence of eligible area for territorialized agri-environmental measures (TAEM) for water issue Share or presence of area in Natura Natura2000 Dummy 2000 Presence or number of certified farms for a Product Conformity Certificate (PCC) Presence or number of certified farms for a Red Label (RL) Presence or number of certified farms for another official label Presence or number of certified farms for a PDO (except wine) Presence or number of certified farms for a wine PDO

11

MONO071

Dummy

shar ate12 07

Continuous

lfa3c12

Dummy

29%

ODR

lfa3c34

Dummy

19% 11%

ODR ODR

shar eliDCE

Continuous

20%

0.03 (0.11)

ASP/ODR

shar natura2000

Continuous

71%

0.10 (0.17)

ASP/ODR

0.08 (0.13)

MSA

fa pccf1

Dummy

nbfa pcc

Continuous

67%

0.03 (0.04)

RA 2000

fa rlf1

Dummy

nbfa rl

Continuous

73%

0.04 (0.07)

RA 2000

fa othf1

Dummy

nbfa oth

Continuous

84%

0.09 (0.13)

RA 2000

fa pdo nowinf1

Dummy

nbfa pdo nowin

Continuous

28%

0.03 (0.10)

RA 2000

fa pdo winf1

Dummy

nbfa pdo win

Continuous

29%

0.08 (0.20)

RA 2000

shar dirsal shar process

Continuous Continuous

0.18 (0.14) 0.08 (0.12)

RA 2000 RA 2000

shar wCuma

Continuous

0.11 (0.11)

RA 2000

Please cite this article in press as: Allaire, G., et al., Territorial analysis of the diffusion of organic farming in France: Between heterogeneity and spatial dependence. Ecol. Indicat. (2015), http://dx.doi.org/10.1016/j.ecolind.2015.03.009

G Model

ARTICLE IN PRESS

ECOIND-2343; No. of Pages 12

G. Allaire et al. / Ecological Indicators xxx (2015) xxx–xxx

12

Description

Presence of farms with tourism or catering activity Length of time of OF in the micro-territories Presence of OF certified downstream operators in 2006 Presence or number of beneficiaries for RDP1 measures – type 1 (agri-environment and incomes) Presence or number of beneficiaries for RDP1 measures – type 2 (agri-environment targeted support) Presence or number of beneficiaries for RDP1 measures – type 3 (forest) Presence or number of beneficiaries for RDP1 measures – type 4 (modernization) Presence or number of beneficiaries for RDP1 measures – type 5 (food processing industries) Presence or number of beneficiaries for RDP1 measures – type 6 (training and monitoring) Presence or number of beneficiaries for RDP1 measures – type 7 (rural, non-agricultural) Location quotient of areas in OF within all agricultural land in the department compared to France in 2006

Variable for the spatial probit model (Model I – extension)

Variable for the negative binomial truncated model (Model II – intensity)

Descriptive statistics for the complete sample (unlog-transformed and uncentred variables)

Name

Name

Type

Dummy variable (share of positive)

Tourcater

Continuous

Type

Source

Continue variable average (standarddeviation) 0.42 (0.49)

RA 2000

7.21 (4.51)

INAO/ASP/ODR

timeOF07bin

Dummy

timeOF07

Continuous

79%

dwstrof20061

Dummy

dwstrof20061

Dummy

47%

indic typo1 1

Dummy

rdp1farm typo1

Continuous

84%

rdp1farm typo2

Continuous

rdp1farm typo2

Continuous

indic typo3 1

Dummy

rdp1farm typo3

Continuous

rdp1farm typo4

Continuous

rdp1farm typo4

Continuous

indic typo5 1

Dummy

indic typo5 1

Dummy

26%

indic typo6 1

Dummy

rdp1farm typo6

Continuous

52%

0.02 (0.05)

ASP

indic typo7 1

Dummy

rdp1farm typo7

Continuous

72%

0.02 (0.02)

ASP

LQdepfce sup006

Continuous

1.17 (1.01)

Agence Bio/ODR

References Agence Bio, 2012. L’agriculture biologique franc¸aise: les chiffres clés – Edition 2012 (Chiffres 2011). Allaire, G., Cahuzac, E., Poméon, T., Simioni, M., 2014. Approche spatiale de la conversion à l’agriculture biologique: les dynamiques régionales en France. Econ. Rurale 339–340, 9–31. Allaire, G., Cahuzac, E., Maigne, E., Poméon, T., 2015. Localisation de l’agriculture biologique et accès aux marchés. Rev. Agric. Environ. Stud. 96 (2), 9–31. A.N.D.-I., 2008. Evaluation ex post du Plan de Développement Rural National. Soutien à l’agroenvironnement. Annexe: Etude de cas sur l’agriculture biologique, Marché CNASEA. Anselin, L., 1988. Spatial Econometrics: Methods and Models. Kluwer Academic, Dordrecht. Beauchesne, A., Bryant, C., 1999. Agriculture and innovation in the urban fringe: the case of organic farming in Quebec, Canada. Tijdschr. Econ. Soc. Geogr. 90 (3), 320–328. Bichler, B., Häring, A.M., Dabbert, S., Lippert, C., 2005, September. Determinants of Spatial Distribution of Organic Farming in Germany. Researching Sustainable Systems, Adelaide, Australia, pp. 21–23. Bjorkhaug, H., Blekesaune, A., 2013. Development of organic farming in Norway: a statistical analysis of neighbourhood effects. Geoforum 45, 201–210. Boivin, N., Traversac, J.B., 2011. Acteurs et agriculture biologique dans la fabrique alternative des espaces: le cas de l’Ile de France. Norois 218 (1), 41–55. Cameron, A.C., Trivedi, P.K., 2013. Regression Analysis of Count Data, 2nd ed. Cambridge University Press, Cambridge. Eades, D., Brown, C., 2006. Identifying Spatial Clusters within U.S. Organic Agriculture, Research paper 2006–2010. Regional Research Institute, West Virginia University. Frederiksen, P., Langer, V., 2004. Localisation and concentration of organic farming in the 1990 – the Danish case. Tijdschr. Econ. Soc. Geogr. 95 (5), 539–549. Gabriel, D., Carver, S.J., Durham, H., Kunin, W.E., Palmer, R.C., Sait, S.M., Stagl, S., Benton, T.G., 2009. The spatial aggregation of organic farming in

81%

INAO 0.27 (0.33)

ASP

0.17 (0.15)

ASP

0.06 (0.27)

ASP

0.20 (0.12)

ASP

ASP

England and its underlying environmental correlates. J. Appl. Ecol. 46, 323– 333. Hilbe, J.M., 2011. Problems with zero counts. In: Negative Binomial Regression, 2nd ed. Cambridge University Press, pp. 346–386 (Chapter 11). Ilbery, B., Holloway, L., Arber, R., 1999. The geography of organic farming in England and Wales in the 1990. Tijdschr. Econ. Soc. Geogr. 90 (3), 285–295. Ilbery, B., Maye, D., 2011. Clustering and the spatial distribution of organic farming in England and Wales. Area 43 (1), 31–41. Lesage, J., Pace, R.K., 2009. Limited dependant variable spatial models. In: Introduction to Spatial Econometrics. CRC Press Taylor & Francis Group, pp. 279–321 (Chapter 10). Lewis, D., Barham, B., Robinson, B., 2011. Are there spatial spillovers in the adoption of clean technology? The case of organic dairy farming. Land Econ. 87 (2), 250–267. Mahé, T., 2012. Perspectives en Agriculture Biologique à L’horizon 2015, Veille n◦ 55. Centre d’Etudes et de Prospective. Noe, E., 2004. Does instrumentalization of organic farming lead to enhancement or dissolution? A case study of the local dissemination processes of organic farming. In: XI World Congress of Rural Sociology. Nyblom, J., Borgatti, S., Roslakka, J., Salo, M.A., 2003. Statistical analysis of network data – an application to diffusion of innovation. Soc. Netw. 25, 175–195. Pohl, A., 2009. How do European rural development programmes support organic farming? In: IFOAM–EU Group. Risgaard, M.L., Frederiksen, P., Kaltoft, P., 2007. Socio-cultural process behind the differential distribution of organic farming in Denmark. Agric. Hum. Values 24 (4), 445–459. Sanders, J., Stolze, M., Padel, S., 2011. Use and Efficiency of Public Support Measures Addressing Organic Farming. Thünen-Institute of Farm Economics, Braunschweig. Schmidtner, E., Lippert, C., Engler, B., Häring, A.M., Aurbacher, J., Dabbert, S., 2012. Spatial distribution of organic farming in Germany: does neighbourhood matter? Eur. Rev. Agric. Econ. 39 (4), 661–683.

Please cite this article in press as: Allaire, G., et al., Territorial analysis of the diffusion of organic farming in France: Between heterogeneity and spatial dependence. Ecol. Indicat. (2015), http://dx.doi.org/10.1016/j.ecolind.2015.03.009