Organic controls in Germany is there a need to harmonize?

Transcription

1 Organic controls in Germany is there a need to harmonize? Alexander Zorn 1, Christian Lippert 2, Stephan Dabbert 2 1 Farm Management group of Agroscope, CH-8356 Ettenhausen, Switzerland (alexander.zorn@agroscope.admin.ch) 2 Institute of Farm Management (410a), Universität Hohenheim, D Stuttgart, Germany (christian.lippert@uni-hohenheim.de, stephan.dabbert@uni-hohenheim.de) Poster paper prepared for presentation at the EAAE 2014 Congress Agri-Food and Rural Innovations for Healthier Societies August 26 to 29, 2014 Ljubljana, Slovenia Copyright 2014 by Alexander Zorn, Christian Lippert and Stephan Dabbert. All rights reserved. Readers may make verbatim copies of this document for non-commercial purposes by any means, provided that this copyright notice appears on all such copies.

2 Abstract Organic food markets substantially rely on a reliable control system. To identify and isolate the effect of different factors potentially influencing the uniform implementation of organic controls, we apply a stepwise estimation of different logit models. Using control data of German farms from five important control bodies we identify, first, risk factors for noncompliance of farms, second, the impact of the control body as well as, third, the potential impact of governmental institutions which are in charge for the implementation of the control system. The results indicate a need for a more harmonized implementation of the German organic control system. 1. Introduction Organic production and labelling is regulated by Council Regulation (EC) No 834/2007 (EC, 2007). The implementation of this regulation falls to the member states. The majority of the member states implements a system of approved private control bodies. This also holds for Germany, the most important organic market in Europa and the country with the highest number, namely 20 control bodies (CBs) (European Commission, 2013). Pooling data from five important German CBs, this paper complements earlier studies for different countries by Gambelli, Solfanelli and Zanoli (2014), Zorn, Lippert and Dabbert (2013), Lippert, Zorn and Dabbert (2014) which showed for single CBs in different European countries that non-compliance can at least partly be explained by farm characteristics. Here, we extend the analysis to the influence of CBs and national or regional competent authorities on control results. In Germany, the organic farming legal framework requires that the selection of operations to be inspected additionally to the regular yearly control shall be based on the risk of non-compliance with the organic regulation s requirements. Both, the European regulation and the national German law specify on how to perform such risk analysis. Compulsory criteria are amongst others, the size of the operation, the complexity of the operation, parallel production (organic and conventional) and former non-compliances. 2. Analytical framework and hypotheses According to the Economics of Crime approach (Becker, 1968), an opportunistic organic farmer balances the expected net benefits of complying with the organic standard compared to a situation of non-compliance. The compliance decision of an organic farmer is basically explained by the following equation: ( ) (1), with B. = net benefit; t= time period (year); i = farm (i = 1,, n); NC = non-compliance (NC it = 1 if farmer i does not comply at time t); C = compliance cost saved when infringing upon the standard; P d = probability that a non-compliance is detected; P s = probability of being sanctioned in case of detection; F = fine (e.g., monetary losses); L = present value of future profit losses due to marketing restrictions and damaged reputation; ε it = error term also reflecting individually different preferences such as the personal commitment to organic farming (Lippert et al., 2014: p.2). In case of B > 0, an opportunistic farmer decides not to comply (NC it = 1), otherwise she/he complies with organic farming requirements (NC it = 0). The cost of compliance and possible savings are expected to be higher for unexperienced organic farmers (who face higher information costs regarding compliance), in case of intensive crops (e.g., fruits which require high capital and labour inputs), complex production 1

3 processes (e.g., parallel conventional farm production or organic processing in parallel to organic farming), and livestock farming where specific requirements have to be met. The latter is particularly relevant for pig and poultry farming. Relatively low costs are assumed for committed farmers (e.g., long-standing members of an organic farmers association) and extensive areas of production, such as permanent grassland or sheep and goat husbandry. For some variables, such as the size of the farm or the personality of the farmer, two-sided hypotheses are formulated, as the net benefit of non-compliance depends on the individual weighing of several influential factors. Regarding the influence of single CBs and the effect of the location (i.e. the federal state s authority in charge of supervising the CBs), on the control results, similarly, two-sided hypotheses are formulated. Differences between CBs and competent authorities could result from (unintended) different interpretations and implementations of the organic regulation or conscious rigidity (laxness) against specific types of non-compliance. 3. Method and data In the following we use severe sanctions as proxy variable for the occurrence of noncompliance. The system of seven types of sanctions applied in Germany allows a clear distinction between minor and severe sanctions. In our analysis, the dependent variable equals 1 in case at least one severe sanction has been imposed on the farm during the respective year and is 0 otherwise. Binary choice models are used to test the effect of the farm production characteristics and the factors mentioned in the preceding section. A latent, unobservable variable y* (e.g., the net benefit of not complying with the organic regulation) is considered to depend on the observed variables x i (that influence C t, P d and L in condition (1)), such as the farm size or the farm production characteristics: y* = β 0 + β 1 x β n x n + ε. (2) The larger y*, the higher is the probability P(y = 1) that the observed binary variable y is equal to one. Assuming a logistic distribution of the error term ε, we estimate binary logit models of the following form (Long and Freese, 2006): 1 P( y 1 x,..., x n ) x x 1 ( n n) 1 e Significantly positive β i indicate an increasing effect on the sanction probability by the relevant variable, whereas negative β i document the opposite effect. For each year of analysis, we estimate separate models. All models, firstly, are estimated as full (unrestricted) models covering all potential explanatory variables and, secondly, as restricted models by stepwise excluding non-significant variables from the estimation. To isolate the influences of farm characteristics, of the location (hypothesising a main influence by the competent authority of the federal state on control results) and of the CB, a basic model is stepwise extended by dummy variables. The data originates from five important German organic CBs that provided their complete 2009 and 2010 control data base on organic farms. The sample represents in each year of analysis more than two-thirds of the German organic farms; descriptive statistics are given in table 1. On average, a farm in the dataset is controlled 1.22 times a year. The mean control frequency however differs between the CBs in a range from 1.14 up to 1.29 controls. These differences could result from different implementations by the CBs but also from different risk classification of the clients; if the client structure or the risk criteria differ considerably between CBs, also the control frequency could differ. (3) 2

4 Table 1. Descriptive statistics of the dataset: Number of farms, control and sanction frequency, characteristics of the farm and farm production by control body and for the complete dataset. Control body (CB) (A E) Attribute (Share of farms in the dataset unless otherwise specified) A B C D E Total (dataset) Number of farms, ,891 2,173 6,925 1, ,196 Share in all organic farms, 2010 (BLE, 2011) 17.7 % 9.9 % 31.6 % 6.7 % 3.3 % 69.3 % Number of farms, both years 7,463 4,309 13,327 2,918 1,408 29,425 Number of controls per year (mean) Farms with a severe sanction 2.9 % 4.6 % 9.3 % 4.1 % 8.6 % 6.4 % Farms in 2010, which were severely sanctioned in % 3.8 % 9.3 % 3.6 % 9.7 % 6.0 % Organic control experience (years, with this CB) Processing (control area B) 5.0 % 9.0 % 16.8 % 37.8 % 47.8 % 16.2 % Processing under contract (control area D) 14.1 % 25.6 % 31.9 % 37.3 % 33.0 % 27.1 % Farms which are member in an organic farming association 1.3 % 66.0 % 72.9 % 56.3 % 47.4 % 50.9 % Agricultural area (ha, mean) Farms with Conversion area 19.1 % 32.5 % 47.9 % 52.0 % 58.8 % 39.3 % Conventional area 20.1 % 8.2 % 29.9 % 1.4 % 2.6 % 20.1 % Cereals 19.3 % 43.3 % 46.3 % 43.9 % 23.4 % 37.7 % Root crops 4.6 % 12.3 % 20.0 % 25.6 % 8.5 % 15.0 % Industrial crops 2.1 % 4.1 % 6.7 % 6.2 % 4.0 % 5.0 % Fodder crop production 20.5 % 42.4 % 49.0 % 54.3 % 25.9 % 40.2 % Other arable crops 0.8 % 1.9 % 1.7 % 1.2 % 0.4 % 1.4 % Fresh vegetables 5.4 % 11.9 % 22.6 % 20.0 % 12.4 % 16.0 % Permanent grassland 93.8 % 87.6 % 86.4 % 82.8 % 57.6 % 86.7 % Fruits and berries 3.1 % 4.7 % 7.6 % 15.0 % 11.3 % 7.0 % Grapes 2.1 % 1.0 % 2.3 % 9.3 % 38.4 % 4.5 % Other permanent crops 0.8 % 4.0 % 9.4 % 24.0 % 28.9 % 8.8 % Mixed orchards ( Streuobst ) 38.2 % 27.8 % 25.1 % 28.8 % 13.6 % 28.7 % Bovine animals 39.8 % 41.6 % 50.9 % 60.8 % 30.7 % 46.8 % Pigs 3.8 % 7.7 % 12.3 % 15.6 % 5.8 % 9.5 % Sheep 5.9 % 5.9 % 12.2 % 13.4 % 9.4 % 9.7 % Goats 4.4 % 2.3 % 9.2 % 9.5 % 3.3 % 6.7 % Poultry 6.9 % 11.3 % 22.4 % 16.3 % 9.4 % 15.6 % Bees 1.6 % 1.2 % 3.7 % 3.3 % 0.2 % 2.6 % Sources: Data from five important German control bodies (CBs), Bundesanstalt für Landwirtschaft und Ernährung (2011). 4. Results and discussion Selected model results are shown in table 2. Starting point of the analysis is a model based on farm characteristics only (see the lower part of variables illustrated in table 1). This model 3

5 Basic farm model ( ) Basic farm model (year 2010) with lagged severe sanctions Basic farm model ( ) Basic farm model (year 2010) with lagged severe sanctions Basic farm model & dummy variables for CB B & federal state BW Basic farm model & dummy variables for CB C & federal state NI Table 2. Results of selected logit models explaining the occurrence of at least one severe sanction. Model type Unrestricted models Restricted models Model specified Observations Bayesian Information Criterion (BIC) McFadden's R² (Pseudo-R²) Agricultural area (ha) *** * *** * ** *** Organic control experience (years) ** ** Processor (yes=1) a Contract processor (yes=1) * * * * Farm is controlled for private organic standards (yes=1) *** *** *** *** *** *** Farm is controlled for international organic standards (yes=1) * * * * * * Conversion area (yes=1) *** ** *** ** *** *** Conventional area (yes=1) *** *** *** *** *** ** Cereals (yes=1) Root crops (yes=1) Industrial crops (yes=1) Fresh vegetables (yes=1) ** * ** * * * Fodder crop production (yes=1) * ** * ** *** Other arable crops (yes=1) Permanent grassland (yes=1) *** ** *** ** *** ** Fruits and berries (yes=1) * * * * Grapes (yes=1) *** ** *** ** *** ** Other permanent crops (yes=1) *** * *** * *** *** Bovine animals (yes=1) *** *** *** *** *** *** Pigs (yes=1) *** * *** ** *** *** Sheep (yes=1) Goats (yes=1) Poultry (yes=1) *** *** *** *** ** ** Bees (yes=1) ,157 13,821 29,157 13,821 29,157 Dummy var. for the year 2010 (yes=1) ** X ** ** ** Mixed orchards ( Streuobst, yes=1) *** *** *** *** ** ** Dummy variable for severe sanction in previous year (yes=1) X *** X *** X X Dummy variable for control body (CB) B X X X X *** X Dummy variable for CB C X X X X X *** Dummy v. for Baden-Württemberg (BW) X X X X *** X Dummy v. for Niedersachsen (NI) X X X X X *** Constant *** *** *** *** *** *** Source: Own calculations based on data from five important German control bodies (CBs). a the variables indicated by yes=1 are dummy variables which take the value 1 if the attribute applies; X this variable was not used in this model. Variables with empty cells, were excluded from the restricted models. Significance levels: * < 0.1, ** < 0.01, *** < ,

6 then is extended by adding variables for former sanctions, for the CB, and, for large federal states. Finally, models for combinations of these extensions are estimated. Such models are estimated for a pooled dataset (covering both years jointly) and for each year separately. Some results are astonishing and contradict preliminary theoretical considerations. This is the case for the sanction decreasing effect of contract processors and grapes, both considered as complex and challenging activities regarding documentation and production. Also, the highly significant effect of farms that adhere to an additional, stricter private standard is not expected. The sanction probability increasing effect of agricultural area may be explained by higher complexity. Extending the model by combinations of dummy variables for federal states and CBs, the comparison of the BIC with the basic model mostly does not indicate a considerable model improvement; nonetheless, the coefficients are nearly all significant. For two of the six federal states with at least observations, we find a highly significant negative coefficient whereas for two other states we obtained significant positive coefficients. In principle, these effects could be due to different natural conditions. However, as we believe that we sufficiently controlled for these conditions by including farm production characteristics (see table 1) into our analysis this result hints at differences in the implementation between the federal states. The same argument applies for the models considering a dummy variable for the CBs; three CBs feature a sanction probability decreasing effect and two an increasing one. The Pseudo-R² values finally obtained are quite low, which is a clear sign that the models are lacking relevant explanatory variables. Thus, future analysis should consider further variables to better represent farmers personal characteristics. We attribute the high relevance of former sanctions partly to the personal characteristics of the farmer. Generally, control data feature the problem of an unknown dark figure of undetected noncompliance. Furthermore, the data could have been influenced by the application of a riskbased control system by the CB. Nevertheless, this analysis for the first time gives evidence on the influence of competent authorities and CBs on organic control results in Germany. This is an important finding, as a credible organic control system is pivotal for a well-functioning organic market. Our results support the need for a more harmonized implementation of organic control systems. References Becker, G. S. (1968). Crime and Punishment: An Economic Approach. Journal of Political Economy 76: Bundesanstalt für Landwirtschaft und Ernährung (2011). Strukturdaten zum Ökologischen Landbau für das Jahr Bonn: BLE. European Commission (2013). List of control bodies or control authorities in charge of controls in the organic sector provided for in article 35(b) of Council Regulation (EEC) No 834/2007. Control Bodies and Control Authorities approved on 31/12/ updates 5 July Brussels: European Union. Gambelli, D., Solfanelli, F. and Zanoli, R. (2014). Feasibility of risk-based inspections in organic farming: results from a probabilistic model. Agricultural Economics 45: Lippert, C., Zorn, A. and Dabbert, S. (2014). Econometric Analysis of Non-compliance with Organic Farming Standards in Switzerland. Agricultural Economics 45: Long, J. S. and Freese, J. (2006). Regression Models for Categorical Dependent Variables Using Stata. Texas: College Station. Zorn, A., Lippert, C. and Dabbert, S. (2013). An analysis of the risks of non-compliance with the European organic standard: a categorical analysis of farm data from a German control body. Food Control 30: