Environmental Modeling & Assessment

, Volume 18, Issue 1, pp 57–72 | Cite as

Evaluating the Efficiency of a Uniform N-Input Tax under Different Policy Scenarios at Different Scales



Nitrate pollution from agriculture is an important environmental externality, caused by the excessive use of fertilizers. The internalization of this problem, via a tax on mineral nitrogen, could lead to a second best solution, reducing nitrate emissions. Several authors suggest that a reduction in agricultural support could produce similar results. In this paper, we examine the effects of different levels of a uniformly implemented nitrogen tax in France under two policy scenarios, corresponding to post Agenda 2000 and 2003 Luxembourg reforms of European Union’s Common Agricultural Policy, in order to reveal the synergies and conflicts between the tax and the policy scenarios in terms of nitrate emissions abatement. The analysis is performed at different geographical scales, from the national to the regional and is based on a bioeconomic approach that involves the coupling of the economic model AROPAj with the crop model STICS. Results show that the efficiency of the N-tax varies according to the geographical scale of the analysis and the type of farming. Furthermore, we prove that a uniform implementation may lead to perverse effects that should always be taken into account when introducing second-best instruments.


Bioeconomic model Mathematical programming Nitrogen response curves Nitrate emissions Nitrogen tax 

1 Introduction

The evolution of the agricultural sector in the European Union (EU) during the second half of the twentieth century has been accompanied by numerous environmental problems, one of which is the pollution of underground and surface waters with nitrates. Nitrates occur naturally in the soil as products of the nitrogen cycle (nitrification process), but also have anthropogenic origins, the most important of which is the application of inorganic or organic fertilizers that directly add both ammonium and nitrate ions to the soil nitrogen pool. The increase of livestock density, which leads to a considerable production of liquid manure waste per area unit of land, has also contributed to the increase of nitrate concentrations in soils.

From an economic perspective, the existence of nitrates in soils reveals a trade-off effect: they serve as nutrients for crops, but at the same time they are pollutants, producing a negative externality, which leads to a social welfare loss that is not compensated for and prevents the attainment of a Pareto optimal allocation of resources. When emissions are measurable, the implementation of either a Pigouvian tax on the effluent itself or appropriate emission quotas can set pollution at the socially desired level and eliminate this externality. In the case of nitrates, which are a typical case of nonpoint source pollution, individual emissions are diffused and cannot be measured precisely, due to technical or economic limitations that hinder the use of the traditional emission-based policy instruments. More specifically, nitrates have both natural and anthropogenic origins that are not easily distinguished and furthermore, the amount of nitrates produced in soils varies locally within the same region; this spatial heterogeneity of nitrate production means that the same management for the same crop in different fields will not necessarily lead to similar nitrate losses. Additionally, losses do not depend only on the soil type but also on the weather conditions, which implies that random factors influence farmers’ contribution to pollution. Hanley [1] also notes the great time lag between nitrate losses from a farmer’s field and the consequent pollution of water bodies, which adds to the uncertainty about the true costs and benefits of any instrument of nitrate pollution control.

Due to the above problems, literature proposes policy instruments that do not deal directly with emissions, but rather focus on taxing or regulating the use of inputs that are responsible for these emissions. Concerning which input should be taxed, inorganic fertilizers, or their nitrogen content, are always the first that come in mind when dealing with nitrate pollution. Taxes increase the marginal cost of the input and consequently reduce its use from a profit maximizing farmer. On the other hand, regulating nitrogen fertilizer use is usually proposed within the best management practices framework and involves both optimizing the application schedule and reducing the amount of fertilizer applied.

The goal of any environmental policy instrument is to guide the economy to a Pareto optimal position, where the level of emissions corresponds to the point where polluters’ marginal benefits are offset by social marginal damages. The problem is that the social damage function is usually either impossible to construct or very costly and difficult to estimate. Therefore, a policy maker’s goal is not social optimality but rather to find ways of achieving desired environmental standards at least cost [1]. Such an approach, which does not necessarily lead to Pareto optimality, but will realize the emission reduction target in the least-cost way, was first presented by Baumol and Oates [2] and is called the “Environmental Pricing and Standards” (EPS) scheme. It involves initially the establishment of a socially acceptable standard of environmental quality and then an iterative set of taxes on either inputs or effluents that affect both outputs and pollution and are continuously readjusted through a “trial and error” procedure, until the environmental target is achieved. The EPS scheme has been widely discussed by numerous authors who identified its shortcomings and potential uses [3, 4].

Concerning the choice of an efficient policy instrument to control nonpoint pollution, Griffin and Bromley [5] utilize the least-cost framework of the EPS approach to examine different abatement policies of input regulation and input management incentives, under the assumption of complete information on emissions and polluters’ returns. They conclude that, when properly specified, all policies are equally efficient and attain the goal of least-cost pollution control. Shortle and Dunn [6] build on the work of Griffin and Bromley and relax the strong assumptions of certainty by treating the runoff process and weather as random variables. They examine different policy options for achieving agricultural nonpoint pollution abatement, namely standards and economic incentives on management practices and estimated emissions. They conclude that although neither policy achieves a first-best solution, management practice incentives are preferable under incomplete information on the runoff process, since they permit the farmer to fully utilize his experience and knowledge of his own farm operations.

Ambient taxes or subsidies have also received considerable attention in the literature. Their conceptual debut is attributed to Segerson [7] who proposed an ambient incentive scheme that can achieve an efficient solution for controlling water quality, based on a Cournot–Nash equilibrium. This approach involves the continuous monitoring of the ambient pollution levels and the implementation of a uniform tax (penalty) for all producers, if the ambient pollution level exceeds the desired standard and a uniform subsidy (credit) when it is lower. Further contributions have examined the moral hazard issue related to ambient pollution based schemes [8] and the role of asymmetric information concerning the fate of effluents [9]. Shortle et al. [10] also identify equity problems that may limit the political acceptability of an ambient instrument, since firms that pollute the most may profit from the abatement efforts of others.

An important issue concerning the selected nonpoint pollution control policy is whether its implementation should be differentiated or uniform. In the case of nitrate pollution, non-uniformity involves regulating or taxing the use of every input for each of the concerned farmers. Obviously, this option may not be feasible, since imposing numerous different charges or quotas implies increased monitoring and enforcement costs. On the other hand, a uniformly implemented tax requires that input use must lead to the same marginal damage across all polluters [11] but in the case of nitrates this is not possible due to the different intensity of input use among farmers and the spatial heterogeneity of nitrate production in soils. Claassen and Horan [12] also raise the issue of equity related to uniform second-best policies. The policy maker’s decision should thus depend on the comparison of the welfare losses of a uniform implementation with the increased costs of the differentiated policy. Helfand and House [13] argue that the former leads only to minor welfare losses compared to the latter, a result which Shortle et al. [10] find “unusual”. Yet, empirical evidence provided by Martínez and Albiac [14] verify the findings of Helfand and House.

Empirical work on the relative performance of the various instruments for controlling nitrate pollution has produced ambiguous results. Gallego-Ayala and Gómez-Limón [15] find that a quota on nitrogen fertilizer is the most cost-efficient instrument, while Semaan et al. [16] find that incentives for better management constitute the least-cost method for reducing nitrate runoff. Taylor et al. [17] argue that no policy is optimal and actual results can vary even among farms within the same region and under the same weather conditions, while similar conclusions are drawn from the paper of Wu and Babcock [18]. Finally, as far as we know, there is no empirical work on ambient taxes or subsidies, only experimental designs [19, 20] that yield different results concerning the efficiency of the ambient instrument.

Other authors also suggest that the elimination of agricultural price support could contribute to nitrate pollution abatement, since it would lead to a reduction in the use of chemical fertilizers because of the lower value of their marginal product. For example, Abler and Shortle [21] show that the elimination of farm commodity programs can lead to a significant reduction in input use, whereas De Haen [22] argues that this cutback in the price support of agricultural products should be significant in order to actually lead to lower fertilizer intensity.

Concerning the various reforms of the Common Agricultural Policy (CAP) in the EU, Wier et al. [23] analyzed the environmental effects of the Agenda 2000 reform and found that, even though inorganic fertilizer use decreases, the reform has only a minimal impact on nitrogen losses, due to the combined effect of changes in animal production, crop mix, and fertilization patterns. It should be noted, however, that Agenda 2000 and most previous reforms only involved a piecemeal decrease in agricultural price support. On the contrary, the 2003 reform of the CAP (introduced in regulation (EC) No. 1782/2003) constitutes an unprecedented and radical shift of support regimes, by introducing full or partial decoupling of subsidies from production. Schmid et al. [24], who examine the environmental effects of the 2003 CAP reform in Austria, conclude that it will have a positive effect, mostly due to land use changes and production output decline, which will in turn lead to a decrease in animal wastes and in chemical input use. They also suggest that the cross compliance measures will have no effect in EU countries already applying agri-environmental programs with stricter standards. On the contrary, Mosnier et al. [25] argue that decoupling can have a positive effect on the environment only when accompanied by the cross-compliance measures. Finally, Gallego-Ayala and Gómez-Limón [15] consider the 2003 CAP reform as a significant instrument for solving the nitrate problem and can be complemented with other policy tools, such as input taxes or quotas, when the reduction in nitrate emissions is not deemed sufficient.

The previous discussion on the different policy instruments and their specificities provides the starting point for this paper, in which we focus on a uniform tax on mineral nitrogen (N-tax), implemented in France, in order to assess its effectiveness and cost-efficiency under changes in the agricultural policy context and at different geographical scales. To this respect, the decoupling of subsidies introduced in the 2003 CAP reform provides an ideal setting to investigate the robustness of an N-tax, i.e., whether such a policy change amplifies or weakens the effects of the tax on nitrate emissions, input use and abatement cost.

What differentiates this work from all the other studies presented previously is that we do not study a tax and a policy reform per se, but we assess their combined effect and identify the synergies or conflicts that arise between them. Additionally, our empirical analysis is not limited to a farm or a regional case study but we examine the different results brought about by the simultaneous implementation of the two instruments at various geographical scales in order to identify shortcomings and inefficiencies that are also due to the uniformity of the N-tax. More precisely, we start from the national scale (France) and then pass to a river basin (the Seine river basin in northern France) and even lower to the regional scale (eight regions that comprise the Seine river basin). Our methodology uses a bioeconomic approach and is based on the coupling of the agricultural supply model AROPAj with the crop growth model STICS. It involves the use of STICS for the estimation of exponential nitrogen response functions, which are integrated in AROPAj, as well as linear functions of nitrate production in soils. An N-tax is then simulated with AROPAj for each policy scenario, by performing 21 simulations that involve increasing the tax level up to 100 % the input price.

2 A comment on Input Taxes

Apart from the theoretical and empirical work on the nonpoint policy options, the implementation of any of the instruments proposed in the previous paragraphs is subject to political and practical constraints that may lead to less efficient solutions than the ones expected in theory. More specifically, regardless of the policy instrument chosen, it is obvious that a uniform implementation is usually a one-way solution since the great number of farmers operating in a single region hinders a differentiated approach. Examining relative efficiency from a policy maker standpoint, a tax is more effective than input usage regulations in terms of monitoring costs, since it is imposed directly on the input price thus changing the value of marginal product of the affected crop, considering of course an exogenous crop price specification. This does not dismiss by any means the option of optimal management practices, including quotas on the fertilizer used. In fact, farming practices that involve optimization of application dates and per hectare quantities are considered as an efficient way of reducing nitrate losses. Yet, there is evidence that a threshold exists beyond which further reductions in fertilizer applications can have positive effects on nitrate runoff only at increasing costs to producers [26]. On the contrary, an input tax provides greater flexibility to the farmer who can adjust his production decisions and thus reduce the on-farm cost of the tax implementation. This is also supported by Shortle and Dunn [6] who argue that in the case of an input tax, the farmer utilizes specialized knowledge on his farm in order to achieve better welfare outcomes than those achieved by quantity regulation instruments imposed by the policy maker. Furthermore, a tax is always related with income redistribution inside the society due to the double-dividend effect that accompanies it.

However, experience has shown that often most actions to control nitrate pollution are based on limiting fertilizer applications rather than taxing its use. This is evident in the EU’s Nitrate Directive (91/676/EEC), which aims at reducing and preventing water pollution by nitrates from agricultural sources and dictates that member States are responsible for identifying pollution sources, designating “vulnerable” zones and designing appropriate action programs. The measures included in each program involve the definition of periods when the application of certain types of fertilizers is prohibited, the definition of proper manure storage methods, the establishment of a volunteer code of good agricultural practices and a mandatory maximum annual per hectare limit of 170 kg of nitrogen from livestock manure. Nevertheless, uniform input taxes are not unknown to European countries. In fact, Finland, Sweden, and Austria introduced a fertilizer levy in 1975, 1985, and 1986 respectively and abolished them before their accession in the EU. Although the tax, which varied in rate with time, resulted in a decrease in fertilizer use, overall agricultural production did not decline, probably due to technological developments that increased fertilizer efficiency. Farmers’ income also remained unchanged since high input prices were offset by export subsidies [27].

The main problem concerning the implementation of an input tax is the uncertainty on the abatement results achieved through limiting input use by increasing its cost. Although for a single crop, the effect of an input tax is straightforward, when a farmer faces multiple crop production possibilities where each one is represented by different patterns of input use (production functions) and contributes differently to nitrate pollution, an input tax could actually lead to completely opposite results than the ones expected in theory. This can be demonstrated by using simple calculus techniques.

Let us assume a farmer’s profit function with I crops (i = 1,2,…,I) and two inputs, nitrogen applied per hectare (N i ) and land of a fixed total amount (X i ) with homogeneous characteristics for each crop:
$$ R\left( {\left. {{X_i},{N_i}} \right|w,p} \right) = \sum\limits_{{i = 1}}^I {{X_i}\left[ {{p_i}{Y_i}\left( {{N_i}} \right) - w{N_i}} \right]} $$

In the above equation, R denotes farmer’s gross margin, w the price of nitrogen, p i the price of crop i, and Y i its yield, expressed as a nitrogen response function that is assumed to be monotonically increasing and concave (Y′ > 0 and Y″ < 0), as is required by both crop science and economic theory. At the optimum, the yield achieved and the amount of nitrogen used will depend on the ratio of nitrogen price to crop price, w/p i . Consequently, the allocated crop area X i will depend on the set of price ratios (w/p 1, w/p 2,… w/p I ).

Production of each crop i is also associated with a function E i (N i ) of per hectare nitrate emissions, which depends on the amount of nitrogen used, with dE i /dN i  > 0. In a specific region, ambient pollution, L, is equal to the sum of nitrate emissions from every activity and is calculated as the product of E i and the area X i occupied by each crop:
$$ L = \sum\limits_{{i = 1}}^I {{X_i}{E_i}} $$
Thus, nitrate emissions depend on the decisions taken both at the extensive margin (crop area) and the intensive margin (nitrogen used). However, although at field level an increase in w will always lead to a decrease in the amount of nitrogen, this may not be the case for the area, X i , allocated to each crop. In fact, due to the different degree of change in crops’ relative profitability, brought about by the increase in the price of nitrogen, an activity substitution may take place. This means that the derivative ∂X i /∂w can take any sign, depending on the direction of changes at the extensive margin for crop i. To show this, we set for simplicity reasons i = 1,2 so that the aggregated nitrate emission is L = X 1 E 1 + X 2 E 2. Assuming that the previous equation represents the result of a farmer’s profit maximization problem under a land constraint, an increase in the price of nitrogen, w, will result in a change of L, given by:
$$ \frac{{{\text{d}}L}}{{{\text{d}}w}} = \left[ {\left( {\frac{{\partial {X_1}}}{{\partial w}}} \right){E_1} + {X_1}\left( {\frac{{{\text{d}}{E_1}}}{{{\text{d}}{N_1}}}\frac{{\partial {N_1}}}{{\partial w}}} \right)} \right] + \left[ {\left( {\frac{{\partial {X_2}}}{{\partial w}}} \right){E_2} + {X_2}\left( {\frac{{{\text{d}}{E_2}}}{{{\text{d}}{N_2}}}\frac{{\partial {N_2}}}{{\partial w}}} \right)} \right] $$
If an activity substitution takes place due to the change of their relative profitability, the land constraint defines that the decrease of X 1 will be equal to the increase of X 2 so that ∂X 1/∂w + ∂X 2/∂w = 0 and by rearranging terms:
$$ \frac{{{\text{d}}L}}{{{\text{d}}w}} = \left[ {{X_1}\left( {\frac{{{\text{d}}{E_1}}}{{{\text{d}}{N_1}}}\frac{{\partial {N_1}}}{{\partial w}}} \right) + {X_2}\left( {\frac{{{\text{d}}{E_2}}}{{{\text{d}}{N_2}}}\frac{{\partial {N_2}}}{{\partial w}}} \right)} \right] \times \left[ {\frac{{\partial {X_2}}}{{\partial w}}\left( {{E_2} - {E_1}} \right)} \right] $$

The first part in brackets of Eq. (1) represents the results in nitrate pollution caused by intensive margin changes and the second the results due to extensive margin changes. The former is always negative, since X i  > 0, dE i /dN i  > 0 and ∂N i /∂w < 0, implying that an increase in the price of nitrogen will always result in lower nitrate emissions from a single field, when no activity substitution is considered. However, the sign of the second bracketed part cannot be defined, as it depends on the per hectare nitrate emissions of each crop at the optimal nitrogen use level. If crop 2 pollutes more than crop 1 (E 2 > E 1), the sign is positive (we have assumed that ∂X 2/w > 0) and this activity substitution leads to an increase in nitrate emissions. Consequently, the direction of changes in the ambient pollution level will depend on whether the changes at the intensive margin are more significant than the ones at the extensive margin: dL/dw will be negative in the first case and positive in the second.

3 Modeling Nitrate Pollution Control Policies

3.1 Bioeconomic Models

The majority of empirical work on modeling instruments for nitrate pollution control relies on farm-level, regional or social welfare bioeconomic models, based on mathematical programming (MP) and more precisely, static linear programming (LP), nonlinear programming (NLP), or dynamic programming specifications. MP models constitute an approach consistent with microeconomic theory, which is the maximization of income under constraints concerning the availability of fixed inputs, while at the same time they offer a quantitative representation of production technology.

A farm bioeconomic model can be described as a model that integrates agronomic information inside formulations that describe farmers’ decisions on managing resources for producing outputs and associated externalities [28]. The agronomic requirements of a bioeconomic model are covered by specialized crop growth models that can relate crop yields, soil characteristics, and input usage and are the result of the latest advances in agronomy, soil, and crop science. However, when the goal is to infer agronomic results to a regional level, such models seem too unwieldy to handle and are site-specific, using microclimatic conditions and soil data for specific fields. This means that they lack an economic dimension, as they cannot be applied at a larger geographical scale without serious assumptions on the physical data used.

To overcome this problem and to link an agronomic model with an economic model, a common solution is to use the simulation results from the agronomic model in order to estimate response functions that relate yields and runoffs with the factors of production under consideration, ceteris paribus. The derived response functions can then be directly incorporated inside an economic model based on mathematical programming. This kind of coupling economic and agronomic models allows for a realistic and detailed representation of the functional relation between input use and effluent emissions and improves the analysis of the implementation results of any of the policy instruments previously described. Examples of this approach in the study of nitrate pollution abatement policies include among others Helfand and House [13], Larson et al. [29], and Martínez and Albiac [14]. In this paper, we follow the same approach and use the crop growth model STICS in order to estimate nitrogen yield response and nitrate emission functions to be incorporated in the AROPAj model.

3.2 The Economic Model AROPAj

AROPAj is a short-term agricultural supply model that has been widely used for the assessment of the CAP reforms in European agriculture from the farm to the EU scale [30]. It is based on linear and mixed-integer programming and utilizes data from the Farm Accountancy Data Network (FADN), which gives the possibility to expand the utilization of the model in order to include all EU Member States.

Although FADN contains a sample of representative farms for each administrative region at national level, AROPAj maximizes an objective function for farm group types rather than individual farms, each having its own constraint set. This means that AROPAj actually consists of a set of independent models that describe the economic behavior of the corresponding farm type and represent the wide diversity of farming systems encountered in European agriculture, covering most annual crops, grasslands, and major animal production activities found throughout the EU. This farm typology is performed for each administrative region and involves an aggregation (clustering) procedure that utilizes nonhierarchical methods and is based on three farm characteristics: (1) farming type (14 types of farming activities, according to FADN nomenclature), (2) location altitude (<300, 300–600, and >600 m), and (3) economic size, as defined by the economic size unit variable. A detailed discussion about the farm typology in AROPAj can be found in Chakir et al. [31]. One important remark is that for reasons of private data protection, each farm group should be associated with at least 15 farms in the FADN database. The derived farm group types (1,074 in total for EU-151) represent “average” farms of the same type and can be considered homogeneous in terms of farming type, geographical location, and altitude.

The model maximizes gross margin for each farm group type and the activity set concerns crop area and output, animal numbers, animal production (milk and meat) and the quantity of the purchased animal feeds. In total, 32 crop and pasture activities and 31 animal categories (27 for cattle plus one each for sheep, goats, swine, and poultry) are represented in AROPAj. Crop production can be sold at the market price or used for animal feeding purposes (feed grains, forage, and pastures). Cattle categories depend on the age, the gender, the origin (on-farm or bought animals), the final output (dairy or meat), and CAP subsidies. The constraint set includes (1) crop rotation and agronomic constraints, (2) restrictions concerning animal demography and nutritional requirements, (3) restrictions concerning quasi-fixed production factors (land and livestock), and (4) restrictions related to CAP measures. More specifically, crop rotations and agronomic constraints concern limited area allocation and average input use. Demographic constraints for animals define internal relationships between bovine categories. Additionally, some animal categories are considered quasi-fixed capital and hence they are allowed to vary in a limited range in each model run (±15 % of the initial animal numbers). CAP restrictions include production quotas or area limitations, while mutually exclusive discrete choices faced by farmers are modeled with the use of binary or integer variables. De Cara et al. [32] provides a more detailed description of the objective function of AROPAj, the activities, the set of constraints and the shortcomings related to FADN use.

Finally, the calibration of AROPAj is based on the re-estimation of one of the model’s parameters subset through a combination of Monte Carlo and gradient methods, in order to minimize the difference between actual observations and model results for each farm group [33]. The calibrated parameters include animal feeding requirements, grassland yields, and maximal crop area shares.

3.3 Coupling AROPAj with STICS

One important characteristic of AROPAj is its modular structure that allows for the selective use of either technical modules that take into account environmental considerations, or policy modules that include a large range of instruments, such as quotas and taxes on inputs or outputs. The first technical module to be included in AROPAj concerned the estimation of greenhouse gas emissions [32, 33]. Similarly, a specific technical module provides the option of substituting average point yields for crops in each farm group type with a response function of nitrogen. This leads to the transformation of AROPAj from an LP to a NLP model, with respect to nitrogen, since the latter is now regarded as a variable and its optimal use is calculated endogenously, along with the corresponding crop yields. These response functions are the result of a procedure that involves the coupling of AROPAj with the crop model STICS and is described in the next paragraphs.

STICS is a crop growth simulation model, based on water and nitrogen balances and driven by daily climatic data, while utilizing soil characteristics and management practices as inputs. It also consists of a number of modules, each dealing with a different set of biophysical functions, either above the soil (e.g., yield and biomass formation) or beneath it (e.g., water and nitrogen balances). A last module is dedicated to the simulation of management practices [34]. For the present work, the nitrogen balance module is of great importance, as it gives the opportunity to estimate both nitrogen uptake by crops and nitrogen losses that ultimately lead to nitrate formation in the soil.

The coupling of STICS with AROPAj involves the estimation of an exponential nitrogen response function to be incorporated in the latter. This methodology has been presented by Godard et al. [35], but it has never before been used for empirical analysis as the one attempted in this paper. Additionally, we move one step forward and use the output of the nitrogen balance module in STICS to also estimate a linear function of nitrate emissions for every [farm group type-crop] combination in AROPAj. It has to be noted that, besides the exponential one, various specifications of yield functions can also be found in the literature, the most common of which are polynomial ones [13, 14, 29]. All of these functions are concave and increasing in the feasible region of production, implying diminishing marginal returns, thus abiding the characteristics imposed by economic theory. The selection of an appropriate functional form has been the main subject of numerous scientific studies but it is suffice to say that the choice is still controversial [36]. Frank et al. [37] conclude that no functional form should be assumed a priori as there are situations where one would seem preferable over another. On the other hand, nitrate emission functions found in the literature include polynomial (square root and quadratic) [14, 29] or exponential specifications [3].

Beginning with the exponential yield response function, it can be expressed as:
$$ Y(N) = {Y_{{\max }}} - \left( {{Y_{{\max }}} - {Y_{{\min }}}} \right){e^{{ - tN}}} $$
where Y denotes the estimated crop yield, Y max is the maximum attainable yield (under no nitrogen stress), Y min is the minimum yield (with no fertilization), t represents the curvature of the response function and N the quantity of nitrogen applied to the crop (in kg per hectare). The advantage of the selected exponential form is that it has all the necessary attributes of a well-defined production function and at the same time it includes parameters that allow for agronomic interpretation.

The process for estimating nitrogen response functions for every crop in each farm group type in AROPAj comprises two steps. The first step consists on providing various options for soil, weather, and management data for STICS, which leads to a number of data combinations and consequently to an equal number of possible N-response curves. In the second step, a single appropriate response function is chosen from the previous set.

Concerning the first step, the main problem encountered when linking agronomic and economic models is the inconsistency of the data used in each of them. AROPAj utilizes aggregated data for farm group types, including costs (but not quantity) of fertilization, with no geographical reference (apart from the knowledge of the FADN region that they belong). On the contrary, STICS requires field-level information on soil, weather and management practices for each crop, which cannot be found in AROPAj. To overcome this problem, Godard et al. [35] opted for the combined use of a number of European-level databases concerning soil and climate information, in addition to phenology and other crop characteristics:
  • Daily weather data for the year 2002 corresponding to version V2 of AROPAj were retrieved from the Monitoring Agriculture from Remote Sensing (MARS)2 project database; To associate climate data with each AROPAj farm group type, every weather cell was assigned an altitude value by overlaying the grid of the MARS database and that of the digital elevation model of Europe3

  • Soil data for STICS were provided by the European Soil Database4 (ESDB). To identify lands where the modeled crops can be cultivated, the map grid of the ESDB was overlaid with the Coordination of Information on the Environment Land Cover database.5 For every FADN region, the five most common soil types from the ESDB were chosen;

  • Management options for the simulated crops included cultivar type, timing of the crop cycle, irrigation, and nitrogen fertilization. For maize and sunflower, whose cultivars vary greatly with respect to earliness, three varieties, and one sowing date were chosen. For the other crops, one variety and three sowing dates were considered, since their cultivars share practically the same timing crop cycle

  • Two cases of irrigation were opted, namely fully irrigated or rain fed crop

  • The types of chemical fertilizer, the total quantity used, and the number of applications for each crop were based on expert knowledge. Furthermore, application dates were translated into phenological stages instead of calendar dates, in order to ensure an adequate fertilization schedule. In farms with livestock, organic nitrogen input was estimated from FADN, using a per animal head parameter of organic nitrogen and this was simulated through an option in STICS which concerns complementary nitrogen sources (besides chemical fertilizers)

  • Pea crop and a winter wheat were the two possible crops in the preceding year

The sum up of all possible combinations of physical data and management options resulted in 30 or 60 response curves6 for each farm group type and crop. These curves were produced by performing, for each possible data combination, 31 simulations with STICS that involved continuously changing the total quantity of nitrogen applied from zero to a maximum level of 600 kg/ha, using a 20-kg step. Every set of points in the yield–nitrogen space produced by these 31 iterative simulations in STICS, was adjusted to the exponential function represented by Eq. (2), thus yielding 30 or 60 curves with different values for parameters t, Y max, and Y min.

The second step in the process of estimating N-response functions involves the selection of the appropriate response curve among all possible candidates. Initially, the derived curves for each crop were compared with the average yield of the same crop in the selected farm group type in AROPAj and the curves that were below this reference yield were excluded. From the remaining ones, the curve that best satisfied the marginal condition of nitrogen use was finally selected: the value marginal product of nitrogen for crop i at the reference yield level should be equal to its price.
$$ \frac{{\partial {Y_i}}}{{\partial N}}\left( {{N^{*}}} \right) = \frac{w}{{{p_i}}} $$

In the above equation, p i denotes the price of the crop, w the price of nitrogen, ∂Y i /∂N is the marginal product of nitrogen, and N * is the amount of nitrogen applied at the reference yield. In other words, the slope of the response function at the reference yield level should be equal to the ratio of nitrogen price and price of crop i.

The estimation of the nitrate emission function follows that of the yield function. For each of the 31 simulations concerning continuously increasing nitrogen doses, STICS produces not only a yield value, but also values corresponding to nitrogen losses in the form of nitrates7 (NO3–N), ammonia (NH3–N) and nitrous oxide (N2O–N) that result from the simulated cropping activity. These emissions are calculated by the nitrogen balance module in STICS, which is able to simulate the physical processes of nitrification, volatilization, and denitrification that occur in the soil–root system and produce each of these pollutants, respectively. Concerning NO3–N losses, the results from STICS can be adjusted to the following linear function:
$$ e(N) = A \cdot N + B $$
The estimated parameters in (3) are A and B. The former represents the slope of the pollution function, i.e., the marginal contribution of the specified crop to nitrate emissions. Parameter B expresses the quantity of NO3–N that is produced in a specific soil through the physical process of nitrification. Figure 1 presents the relation between yield and the nitrate pollution functions estimated from STICS. It should be noted at this point that the results of STICS concerning NO3–N losses are not linear in all the range of possible nitrogen values. More precisely, linearity is observed up to about 360 kg/ha, after which the emissions become convex. This point is shown by N c in Fig. 1. However, we decide to keep the linear form because 360 kg constitute a realistic upper bound of nitrogen applied to most crops.
Fig. 1

N-yield and nitrate emission functions estimated by STICS

A final note is that the nitrogen balance module in STICS calculates losses of NO3–N only at the root level of the soil–crop system. However, the pollution of water bodies from nitrates constitutes a dynamic procedure that involves a significant time lag between the emission and its polluting effect and, in addition, it is difficult to predict and simulate the actual fate of the nitrate ion after it leaves the upper soil layers, due to the various random parameters that affect it (e.g., weather conditions and soil variability). Such simulations are produced by hydrogeological models that allow for the reproduction of the hydrodynamic behavior of the river basin and simulate the transfer of pollutants on the various soil components. However, these kinds of simulations are beyond the scope of this work, which concerns the efficiency of an input tax to control NO3–N losses. Our methodology provides a static image of the nitrates that are produced in the root zone, as a result of farming activities and the N-cycle, implying that the estimated emissions concern only the geographical location in which they were produced. Therefore, no predictions can be made about the consequent pollution of surface or groundwaters in the region.

3.4 Simulating a Nitrogen Tax in France

In our study, a tax on mineral nitrogen is simulated in France under two policy scenarios, in order to examine how the levy impacts on fertilizer use and NO3–N losses. The two policy scenarios are modeled after the CAP regime at the base year (2002) and the post 2003 CAP reform regime and will be called henceforth “Agenda 2000” and “Luxembourg” respectively. Both scenarios use the same price vectors so that differing results are attributed to the policy changes only. These include mainly the “decoupling” of subsidies for production and the introduction of the single payment scheme for each farm. The Luxembourg scenario also simulates more recent policy changes, namely the abolition of mandatory set aside and the reform of the sugar sector. A detailed presentation of how decoupling is modeled in AROPAj is provided by Galko and Jayet [30]. For every policy scenario, 21 simulations concerning continuously increasing tax levels up to 100 % the N-input price are examined, using a 5 % increase step.

Three geographical levels were chosen for the analysis: (1) the national level (France), (2) the basin level, where we focus on the Seine river basin that consists of eight FADN regions, and (3) the regional level, which corresponds to the FADN regions that comprise the Seine river basin. The latter is located in northern France and covers a surface of about 78,600 km2. It totally encompasses the FADN regions of Île de France and Haute–Normandie and partially those of Champagne–Ardenne, Picardie, Centre, Basse–Normandie, Bourgogne, and Lorraine. Its population amounts about 16 million inhabitants, representing a quarter of France’s total population.

About 15 % of French farms are situated in the Seine river basin area and use 23 % of the total available agricultural land in the country. In terms of farming activities, cereals and protein crops are the dominant crops, taking up about 43 % of the total available agricultural land, while animal production concerns mostly cattle [38]. This information shows that agriculture is the principal activity of the Seine river basin in terms of land use and appears as a major source of pollution of local water bodies with nitrates. In fact, the Seine river basin has been designated a vulnerable zone, as defined by the Nitrate Directive and three different action programs have been implemented since 1996. These programs have focused mainly on the installation of herbaceous zones for the protection of surface waters and on defining a set of good agricultural practices, including intercropping and rules for proper fertilizer use. However, these measures have succeeded only in limiting the increase in aquifer nitrate concentrations, whose median has portrayed an annual increase of 0.64 mg/l over the last 30 years. This clearly shows that a combination of measures to drastically limit nitrogen applications are needed [39]. In this context, a possible tax on mineral nitrogen seems an appropriate scenario that needs to be investigated.

Since the tax is imposed on the nitrogen content of the fertilizer and not on the fertilizer itself, the actual monetary value of nitrogen needs to be estimated. In the simulations performed with STICS, different kinds of fertilizer were used for each crop, which, although desirable from an agronomic point of view, creates difficulties in the modeling of a nitrogen tax: the use of different fertilizers with different prices and different nitrogen contents does not allow for a single answer on the actual value of nitrogen. AROPAj considers a single “average” reference fertilizer for each crop and as the market of nitrogen is hypothetical, we estimated a mean nitrogen price based on the price and content of each reference fertilizer type. More specifically, the price of each reference fertilizer was divided between the relative content of its basic components (N, P, and K), which yielded a mean nitrogen content of 18 % and a mean value of about 1 Euro/kg, depending on the crop. This means that a 100 % N-tax corresponds to a total nitrogen price of 2 Euro/kg.

It should be noted that AROPAj is a static agricultural supply model and thus it does not capture possible market price fluctuations due to the implementation of the N-tax; a tax on nitrogen will reduce demand for fertilizer and will eventually lead to a decrease in fertilizer prices until a new equilibrium is found that may negate the effect of the tax. However, such price variations are not considered in the simulated scenarios because this requires a partial equilibrium approach and thus is beyond the scope of the present work.

As explained previously, a linear nitrate emission function was estimated for every crop in each farm group type, making per hectare NO3–N losses from crops an endogenous variable. For other types of land use (e.g., fallow and grasslands), AROPAj takes account of low intensity nitrogen fertilizer applications and thus the corresponding mineral nitrogen input is exogenous, given by appropriate per hectare parameters. Concerning animal production, to each animal category corresponds a per head parameter of manure (organic N) production, which is then transformed into mineral N equivalent and enters a nitrogen balance in each farm. The latter allows the farmer to choose the source of the N input required for his crops, i.e., either the organic N coming from his animals (transformed into mineral N equivalent), or nitrogen coming from chemical fertilizers. This choice depends on the implicit (shadow) price of on-farm produced N and the market price of mineral N, adjusted by the tax. The 170 kg/ha limitation for organic N from the Nitrates Directive always applies. Hence, nitrate losses coming from organic nitrogen used in crop production are estimated through the emissions function.

4 Results and Discussion

4.1 Results on Nitrogen Fertilizer Use

The implementation of a tax on mineral nitrogen is anticipated to have an impact on both the intensive and the extensive margin. Concerning the former, Table 1 describes the change in fertilizer use at the national and the basin scale as the N-tax increases. More specifically, fertilizer use for different tax levels in each policy scenario is compared to a base case, which corresponds to a zero N-tax under Agenda 2000. The “difference” columns thus represent the percentage difference in fertilizer use between the two scenarios. At the national scale we observe a weak synergistic effect between the tax and decoupling; a tax level of 100 % leads to a decrease in fertilizer use of about 50.5 % under Agenda 2000 and 51.6 % under the Luxembourg scenario. In the Seine river basin, similar synergies appear after an N-tax level of 10 %.
Table 1

Changes in fertilizer use in France and the Seine river basin under different policy scenarios and different N-tax levels

Tax (%)


Seine River Basin

Agenda 2000 (a)

Luxembourg (b)

Difference (b)–(a)

Agenda 2000 (c)

Luxembourg (d)

Difference (d)–(c)














































































aThe value corresponds to 13.26 million tons of fertilizer use

bThe value corresponds to 7.34 million tons of fertilizer use

The above results also show that the demand for fertilizer with respect to the price of its nitrogen content is inelastic, since a percentage increase in the price of nitrogen leads to a lower percentage decrease in fertilizer use. Additionally, because of the hypothesis of fixed N content of fertilizers, the percentage change in nitrogen use at every tax level is identical to that of fertilizer, which in turn implies an inelastic demand for nitrogen. This is generally in line with the existing literature, although the actual fertilizer or nitrogen demand changes due to a tax may vary: For example, Berntsen et al. [40] find that a 100 % N-tax will lead to a reduction of 23–28 %, in nitrogen use, depending on soil type. Compared to our results, this difference can be explained by the fact that their estimated price of nitrogen is 0.67 Euro/kg, which is significantly lower than our 1 euro/kg average. On the contrary, Schou et al. [41] find that a similar N-tax will result in a significant reduction of fertilizer consumption, ranging from 40 to 80 %, depending on the soil type and the farm system examined, although their estimation of the value of mineral nitrogen is even lower than that of Berntsen et al., reaching only 0.42 Euro/kg.

The aggregate picture at the national and the regional scale can be misleading since each region of the Seine river basin differs greatly from the other in terms of fertilizer use. This is clearly shown in Table 2 that presents the combined effects of policy change and of the N-tax in two FADN regions within the Seine river basin area, namely Basse-Normandie (FADN region 135) and Île de France (FADN region 121).
Table 2

Changes in fertilizer use in Basse–Normandie and Île de France under different policy scenarios and different N-tax levels

Tax (%)


Île de France

Agenda 2000 (a)

Luxembourg (b)

Difference (b)–(a)

Agenda 2000 (c)

Luxembourg (d)

Difference (d)–(c)














































































aThe value corresponds to 395,300 tons of fertilizer use

bThe value corresponds to 458,400 tons of fertilizer use

According to Table 2, fertilizer use at all N-tax levels in Île de France increases slightly due to decoupling, while in Basse–Normandie decreases up to a maximum of about 13 % (see the “Difference” column at an N-tax level of 25 %). These results can be explained by examining the production orientation of the farm group types comprising each region: Île de France produces mostly arable crops, with soft wheat, maize, barley, rapeseed and sugar beet taking up about 85 % of the region’s total agricultural land. The N-tax leads to a fall in crop yields (intensive margin change) that is similar in both policy scenarios, while decoupling increases the surface allocated to the previous crops, which now take up about 93–94 % of total land (extensive margin change). Most of these extra hectares come from the abolition of the set aside (subsidized) regime in the Luxembourg scenario. The increase in fertilizer use in Île de France shows that extensive margin changes brought about by the Luxembourg scenario undermine the ability of the N-tax in reducing fertilizer use.

On the other hand, Basse–Normandie produces mainly livestock and at the same time 35 % of total agricultural land is covered by permanent meadows. The Luxembourg scenario leads to a small increase in the livestock units raised and in land allocated to permanent meadows, while land allocated to arable crops decreases marginally. The increase in livestock units also results in an increase of the on-farm produced organic nitrogen which may be used in crop production. Additionally, the N-tax increases the market price of chemical fertilizers and finally reduces its demand. These results lead to a reduction in fertilizer use under the Luxembourg scenario for all N-tax levels.

The above presentation clearly shows that decoupling by itself does not guarantee a reduction in input use, since the actual changes depend on the combination of policies implemented in the 2003 CAP reform, one of which is the abolition of subsidized set aside. Additionally, it is evident that the type of farm group (production orientation) is the most important factor affecting fertilizer use.

4.2 Results on NO3–N Losses

The effect of the N-tax on NO3–N losses at the national and the basin scale are presented in Table 3, where we observe that they follow the changes in fertilizer use; in France, the tax is more effective at reducing nitrate emissions under the Luxembourg scenario. In the Seine river basin, however, the results are slightly different, since similar weak synergistic effects between the tax and decoupling are observed after a tax level of 10 %.
Table 3

Changes in NO3–N losses in France and the Seine river basin under different policy scenarios and different N-tax levels

Tax (%)


Seine River Basin

Agenda 2000 (a)

Luxembourg (b)

Difference (b)–(a)

Agenda 2000 (c)

Luxembourg (d)

Difference (d)–(c)














































































aThe value corresponds to 843,100 tons of NO3-N losses

bThe value corresponds to 258,004 tons of NO3–N losses

For France, a 100 % tax level leads to a decrease of 15.5 % in nitrate emissions under the Agenda 2000 and of 16 % under the Luxembourg scenario. In the Seine river basin, this decrease is estimated at about 17.8 % for the Agenda 2000 and 19.9 % for the Luxembourg scenario. These results show that at these geographical scales nitrate response to an N-tax is inelastic, which is in accordance with previous studies; for example, Shou et al. [41] find that a similar tax will result in a 20–22 % reduction in nitrate leaching, depending on farm type.

Examining the effects of the N-tax on activity levels (extensive margin) and crop yields (intensive margin) at the various geographical scales can shed more light to the analysis of the previous results. For France and the Seine river basin, the changes brought about by a 100 % N-tax under both policy scenarios are presented in Table 4, where we observe that the tax would lead to a significant decrease in cereal production and an increase in set-aside and grasslands. On the contrary, the increase in animal production, mostly under the Luxembourg scenario, seems to limit the effectiveness of the N-tax because it increases the organic nitrogen produced on-farm, which is not affected by the tax and acts as a replacement of mineral nitrogen used in crop production.
Table 4

Results of a 100 % N-tax in activity levels and farm gross margin, in France and the Seine river basin under different policy scenarios



Seine River Basin

Agenda 2000


Agenda 2000


Farm gross margin (%)





Cereal surface (%)





Oleo-protein surface (%)





Forage surface (%)





Permanent grassland (%)





Set Aside and fallow (%)





Livestock units (%)





Cereal production (%)





Average cereal yield (%)





As in the case of fertilizer use, this aggregated picture is the sum up result of changes in each FADN region, which can vary both in magnitude and direction. For example, the implementation of the N-tax in Lorraine (FADN region 151) under the Luxembourg scenario leads to lower nitrate emissions than under Agenda 2000. On the contrary, in Île de France, NO3–N losses are higher under the Luxembourg scenario and most importantly, for both policy scenarios, after an initial slight decrease, the N-tax actually leads to a significant increase in nitrate emissions. This is observed at a tax level of 75 % for Agenda 2000 and of 35 % for the Luxembourg scenario (Table 5).
Table 5

Changes in NO3–N losses in Lorraine and Île de France under different policy scenarios and different N-tax levels

Tax (%)


Île de France

Agenda 2000 (a)

Luxembourg (b)

Difference (b)–(a)

Agenda 2000 (c)

Luxembourg (d)

Difference (d)–(c)




















































































































































aThe value corresponds to 34,960 tons of NO3–N losses

bThe value corresponds to 12,900 tons of NO3–N losses

The reason for this unexpected increase in NO3–N losses lies in the changes brought about by the N-tax in the extensive margin, i.e., the surface allocated to each crop and can be explained by the theoretical point discussed in Section 2. Examining these changes in both policy scenarios, it is clear that the sudden increase in NO3–N losses occurs simultaneously with the significant increase in the area allocated to soft wheat and the reduction in that of maize (Fig. 2). At the same time, land allocation to other arable crops (like rapeseed and barley) remain practically unchanged. This means that the increase in nitrate emissions in both scenarios, despite the implementation of the N-tax, is the result of a “crop substitution effect”, caused by the change in the relative profitability of the farming activities in the region. More specifically, as the N-tax increases, and due to the different yield functions of each crop, the changes in crops’ gross margins lead to an increase in area allocated to the more profitable ones, which, however, contribute more to nitrate emissions than their predecessors. At higher geographical scales, the tax-induced reduction in nitrogen use in these crops is not enough to compensate for the increase in the surface of crop land, resulting in the subsequent increase in nitrate emissions.
Fig. 2

Crop surface area changes under different N-tax levels in Île de France for the Agenda 2000 (a) and the Luxembourg (b) scenario

The above discussion shows that locally, crop substitution, caused by an input tax, may possibly have a more significant effect on NO3–N losses than the reduction of input use. The example of Île de France proves that uniform second-best policies fail to capture the spatial variability in externality generation and for this reason more disaggregated instruments are required. When a non-uniform implementation is opted at the regional scale, the different tax levels chosen for each region must be carefully defined so that no such perverse effects are produced. Thus, in cases like Île de France, the tax should never surpass the point at which the “crop substitution effect” occurs. However, when this point corresponds to a low tax level and considering the inelastic response of nitrates, it is clear that the input tax is unable to achieve a significant reduction in NO3–N losses. One solution for eliminating this substitution effect is a tax on land use, proposed by Goetz et al. [42]. Putting aside any issues of political acceptability, the land use tax could be imposed on the crops that contribute the most to nitrate emissions and complement the N-tax in its abatement role.

4.3 Assessing the Cost-Efficiency of the N-tax and the Policy Scenarios

The result of the N-tax in social welfare is calculated by adding the abatement cost involved in the tax scheme and the welfare gain that is due to the decrease in the social damage caused by the nitrates. The abatement cost is defined as the difference between foregone income from the producers’ side and the amount of money received by implementing the tax. Transaction, administration, control, and enforcement costs are also important, but are not considered in the analysis. Since the decrease in social damage function is impossible to construct, we follow a least-cost framework, according to which, the most cost-efficient policy scenario will be the one with the lowest marginal abatement cost for a given pollution reduction target (i.e., local or regional nitrate reduction).

In empirical studies found in the literature, this reduction target often refers to a desired nitrate concentration in water extracted from an aquifer. For this reason, the cost of a policy instrument for attaining this target is compared to the corresponding unitary cost of denitrification from water treatment plants (e.g., [42, 43]). This method also provides a good proxy of the costs involved in off-site abatement choices and allows their comparison with on-site policy instruments of nitrate pollution abatement. In our case, however, total nitrate emissions are calculated at the root level locally and not at a single catchment area or an aquifer, which means that we cannot compare our model’s results to any unitary denitrification cost. Therefore, in the absence of a specific environmental standard, the assessment of the cost-efficiency of the two policy scenarios should be based on the comparison of their marginal abatement cost functions with respect to different levels of nitrate abatement targets.

We estimated the “total” abatement cost functions by regressing calculated abatement cost (defined previously as foregone income minus tax revenue) on the corresponding nitrate reduction percentage at every tax level. After testing a linear and a quadratic functional form, the latter was selected due to its higher R 2 coefficient.

Figure 3 presents the marginal abatement cost functions for the two policy scenarios in France and the Seine river basin. These functions are assumed to be linear and were produced by the derivation of the corresponding total cost abatement functions. At the national scale and for low abatement targets (below 6 %), the N-tax under Agenda 2000 seems more cost-efficient than under the Luxembourg scenario, while the opposite applies for higher abatement targets. On the contrary, in the Seine river basin, the tax under the Luxembourg scenario is uniformly more cost-efficient than Agenda 2000.
Fig. 3

Marginal abatement cost functions for France (a) and the Seine river basin (b)

Concerning the cost-efficiency of the two policy scenarios at the regional scale, Fig. 4 presents the marginal abatement cost functions for Basse–Normandie and Haute–Normandie (FADN region 134), which were again assumed to be linear.
Fig. 4

Marginal abatement cost functions for Basse–Normandie (a) and Haute–Normandie (b)

As described previously, Basse–Normandie produces mainly livestock, while decoupling leads to a decrease in land allocated to arable crops. Hence, the N-tax has a lower impact on activity levels (and consequently on farmer’s revenue) and additionally, due to the lower use of fertilizers (see Table 2) nitrate reduction under the Luxembourg scenario is achieved at a lower marginal cost than under Agenda 2000. On the contrary, Haute–Normandie produces mainly arable crops, which means that the tax has a greater impact on agricultural production and thus nitrate abatement is attained at increased costs. This analysis reveals that farming differences between each region play an important role on cost-efficiency.

5 Conclusions

Due to the nonpoint source nature of nitrate pollution, input taxes are often used as instruments of nitrate pollution control. In this paper, we study the effects of different levels of an N-tax on nitrate emissions from agriculture in France, using a bioeconomic approach that is based on the coupling of the economic model AROPAj with the crop model STICS. The selected approach takes into account the spatial heterogeneity of nitrate emissions at the root level of crops and provides a significant added value in the modeling of the complex relations pertaining to the crop–soil system. The analysis performed is twofold. Initially, we examine how the efficiency of the N-tax is affected by a change in the policy context, by simulating two different policy scenarios, namely the “Luxembourg” and the “Agenda 2000” scenarios that represent the 2003 reform of the CAP and its preceding regime respectively. Additionally, by comparing the results at three different geographical scales (France—Seine river basin—FADN regions of Seine river basin) we show that uniform input taxes constitute a second best solution that includes considerable uncertainty on the true costs and benefits of the corresponding abatement effort.

At the national scale, the N-tax performs better at reducing nitrate emissions and fertilizer use at any tax level under the Luxembourg scenario. Similar synergies between the tax and the Luxembourg scenario appear in the Seine river basin after a tax level of 10 %. Additionally, the Luxembourg scenario increases the cost-efficiency of the N-tax in the Seine river basin, while the same results are observed for France when nitrate abatement targets are set to more than 6 %.

When we pass to the regional scale, the effects of the N-tax on fertilizer use, nitrate emissions and cost-efficiency under both policy scenarios differ greatly from one region to the other and depend on the prevailing farming activity in each examined region. Results indicate that in regions with arable crops it is more difficult to define the exact direction of change in nitrate emissions, since the decrease in nitrogen use is sometimes overlapped by extensive margin changes that involve an increase in land allocated to more profitable, but at the same time more polluting crops. This is demonstrated in the case of Île de France where a paradoxical increase in NO3–N losses occurs exactly at the tax level where a partial substitution of maize with soft wheat is observed.

A word of caution is needed regarding weather conditions. Our analysis treats regional precipitation patterns as a model parameter and therefore the effect of rainfall on nitrate emissions at the root level is assumed to be constant. Although this does not allow for an investigation of how significant the role of precipitation on the reported level of emissions really is, it does not change the main empirical finding of this work, i.e., that an N-tax can, in some cases, lead to the opposite result from the one which was initially intended. This means that even if the increase of nitrate emissions in Île de France may be caused by a specific rainfall pattern, such a case still remains a possibility that adds to the uncertainty of the true abatement results of uniform second-best policies and thus cannot be ignored by policy makers.

To sum up, our results prove that under some specific regional weather conditions and structural settings, the implementation of a uniform input tax may lead to the amplification of nitrate emissions, whereas in other cases a significant abatement level can be achieved. This means that the N-tax fails to deal with the geographical heterogeneity of nitrate emissions and it highlights the need for policy tools that will provide a balance between abatement level and cost. For example, combining an N-tax with application standards may be more cost-efficient and possibly limit the impact of crop substitution on NO3–N losses. Similarly, a land use tax on the most polluting crops is proposed for eliminating the “crop substitution effect”. Examining the relative efficiency of these two options presented above is beyond the scope of the present work; however both merit further study since they provide a promising framework for increasing the efficiency of an input tax as a means to achieve nitrate pollution abatement.


  1. 1.

    In this study, the V2 version of the model is used, which concerns the version developed for the GENEDEC program and is associated with FADN data for 2002 and EU-15.

  2. 2.

    For more information: http://www.marsop.info/marsop3/.

  3. 3.
  4. 4.

    For more information: http://eusoils.jrc.ec.europa.eu/.

  5. 5.
  6. 6.

    For rain-fed crops, two irrigation options were considered (full irrigation and rain), while for crops with high water requirements only full irrigation was opted. Hence, 60 curves were created for the former and 30 for the latter.

  7. 7.

    For the rest of the paper, we will use the terms “nitrate emissions” and “nitrogen losses in the form of nitrates”, or simply “NO3–N losses” interchangeably.



This paper is based on research activities funded by the PIREN-Seine, an interdisciplinary research program dedicated to the study of the environment in the Seine river basin in France.


  1. 1.
    Hanley, N. (1990). The economics of nitrate pollution. European Review of Agricultural Economics, 17, 129–151.CrossRefGoogle Scholar
  2. 2.
    Baumol, W. J., & Oates, W. E. (1971). The use of standards and prices for protection of the environment. The Swedish Journal of Economics, 73(1), 42–54.CrossRefGoogle Scholar
  3. 3.
    Abrams, L. W., & Barr, J. L. (1974). Corrective taxes for pollution control. An application of the environmental pricing ond standards systems to agriculture. Journal of Environmental Economics and Management, 1, 296–318.CrossRefGoogle Scholar
  4. 4.
    Tietenberg, T. H. (1973). Controlling pollution by price and standard systems: a general equilibrium analysis. The Swedish Journal of Economics, 75(2), 193–203.CrossRefGoogle Scholar
  5. 5.
    Griffin, R. C., & Bromley, D. W. (1982). Agricultural runoff as a nonpoint externality: a theoretical development. American Journal of Agricultural Economics, 64(3), 547–552.CrossRefGoogle Scholar
  6. 6.
    Shortle, J. S., & Dunn, J. W. (1986). The relative efficiency of agricultural source water pollution control policies. American Journal of Agricultural Economics, 68(3), 668–677.CrossRefGoogle Scholar
  7. 7.
    Segerson, K. (1988). Uncertainty and incentives for nonpoint pollution control. Journal of Environmental Economics and Management, 15(1), 87–98.CrossRefGoogle Scholar
  8. 8.
    Xepapadeas, A. P. (1991). Environmental policy under imperfect information: Incentives and moral hazard. Journal of Environmental Economics and Management, 20(2), 113–126.CrossRefGoogle Scholar
  9. 9.
    Cabe, R., & Herriges, J. A. (1992). The regulation of non-point-source pollution under imperfect and asymmetric information. Journal of Environmental Economics and Management, 22(2), 134–146.CrossRefGoogle Scholar
  10. 10.
    Shortle, J. S., Horan, R. D., & Abler, D. G. (1998). Research issues in nonpoint pollution control. Environmental and Resource Economics, 11(3–4), 571–585.CrossRefGoogle Scholar
  11. 11.
    Helfand, G. E., Berck, P., & Maull, T. (2003). The theory of pollution policy. In K. G. Mäler & J. R. Vincent (Eds.), Environmental degradation and institutional responses (handbook of environmental economics, vol. 1) (pp. 249–303). Amsterdam: Elsevier.CrossRefGoogle Scholar
  12. 12.
    Claassen, R., & Horan, R. D. (2001). Uniform and non-uniform second-best input taxes. Environmental and Resource Economics, 19, 1–22.CrossRefGoogle Scholar
  13. 13.
    Helfand, G. E., & House, B. W. (1995). Regulating nonpoint source pollution under heterogeneous conditions. American Journal of Agricultural Economics, 77(4), 1024–1032.CrossRefGoogle Scholar
  14. 14.
    Martínez, Y., & Albiac, J. (2006). Nitrate pollution control under soil heterogeneity. Land Use Policy, 23(4), 521–532.CrossRefGoogle Scholar
  15. 15.
    Gallego-Ayala, J., & Gómez-Limón, J. A. (2009). Analysis of policy instruments for control of nitrate pollution in irrigated agriculture in Castilla y León, Spain. Spanish Journal of Agricultural Research, 27(1), 24–40.Google Scholar
  16. 16.
    Semaan, J., Flichman, G., Scardigno, A., & Steduto, P. (2007). Analysis of nitrate pollution control policies in the irrigated agriculture of Apulia Region (Southern Italy): a bio-economic modelling approach. Agricultural Systems, 94(2), 357–367.CrossRefGoogle Scholar
  17. 17.
    Taylor, M. L., Adams, R. M., & Miller, S. F. (1992). Farm-level response to agricultural effluent control strategies: the case of the Willamette Valley. Journal of Agricultural and Resource Economics, 17(1), 173–185.Google Scholar
  18. 18.
    Wu, J., & Babcock, B. A. (2001). Spatial heterogeneity and the choice of instruments to control nonpoint pollution. Environmental and Resource Economics, 18, 173–192.CrossRefGoogle Scholar
  19. 19.
    Cochard, F., Willinger, M., & Xepapadeas, A. (2005). Efficiency of nonpoint source pollution instruments: an experimental study. Environmental and Resource Economics, 30, 393–422.CrossRefGoogle Scholar
  20. 20.
    Spraggon, J. (2002). Exogeneous targeting instruments as a solution to group moral hazards. Journal of Public Economics, 84, 427–456.CrossRefGoogle Scholar
  21. 21.
    Abler, D. G., & Shortle, J. S. (1992). Environmental and farm commodity policy linkages in the US and the EC. European Review of Agricultural Economics, 19, 197–217.CrossRefGoogle Scholar
  22. 22.
    De Haen, H. (1982). Economic aspects of policies to control nitrate contamination resulting from agricultural production. European Review of Agricultural Economics, 9, 443–465.CrossRefGoogle Scholar
  23. 23.
    Wier, M., Andersen, J. M., Jensen, J. D., & Jensen, T. C. (2002). The EU’s Agenda 2000 reform for the agricultural sector: environmental and economic effects in Denmark. Ecological Economics, 41(2), 345–359.CrossRefGoogle Scholar
  24. 24.
    Schmid, E., Sinabell, F., & Hofreither, M. F. (2007). Phasing out of environmentally harmful subsidies: consequences of the 2003 CAP reform. Ecological Economics, 60(3), 596–604.CrossRefGoogle Scholar
  25. 25.
    Mosnier, C., Ridier, A., Kphaliacos, C., & Carpy-Goulard, F. (2009). Economic and environmental impact of the CAP mid-term review on arable crop farming in South-western France. Ecological Economics, 68(5), 1408–1416.CrossRefGoogle Scholar
  26. 26.
    Johnson, S. L., Adams, R. M., & Perry, G. M. (1991). The on-farm costs of reducing groundwater pollution. American Journal of Agricultural Economics, 73(4), 1063–1073.CrossRefGoogle Scholar
  27. 27.
    >Rougoor, C. W., Van Zeijts, H., Hofreither, M. F., & Bäckman, S. (2001). Experiences with fertilizer taxes in Europe. Journal of Environmental Planning and Management, 44(6), 877–887.CrossRefGoogle Scholar
  28. 28.
    Janssen, S., & Van Ittersum, M. K. (2007). Assessing farm innovations and responses to policies: a review of bio-economic farm models. Agricultural Systems, 94(3), 622–636.CrossRefGoogle Scholar
  29. 29.
    Larson, D. M., Helfand, G. E., & House, B. W. (1996). Second-best tax policies to reduce nonpoint source pollution. American Journal of Agricultural Economics, 78(4), 1108–1117.CrossRefGoogle Scholar
  30. 30.
    Galko, E., & Jayet, P. A. (2011). Economic and environmental effects of decoupled agricultural support in the EU. Agricultural Economics, 42, 605–618.CrossRefGoogle Scholar
  31. 31.
    Chakir, R., Debove, E., De Cara, S., & Jayet, P. A. (2006). The AROPAj model. Work Package 2, Deliverable 2, GENEDEC Project. http://www.grignon.inra.fr/economiepublique/genedec/publi/deliv/WP2_D2_TR.doc. Accessed 9 August 2010.
  32. 32.
    De Cara, S., Houzé, M., & Jayet, P. A. (2005). Methane and nitrous oxide emissions from agriculture in the EU: a spatial assessment of sources and abatement costs. Environmental and Resource Economics, 32, 551–583.CrossRefGoogle Scholar
  33. 33.
    De Cara, S., & Jayet, P. A. (2000). Emissions of greenhouse gases from agriculture: the heterogeneity of abatement costs in France. European Review of Agricultural Economics, 27(3), 281–303.CrossRefGoogle Scholar
  34. 34.
    Brisson, N., Gary, C., Roche, R., Mary, B., Ripoche, D., Zimmer, D., et al. (2003). An overview of the crop model STICS. European Journal of Agronomy, 18, 309–332.CrossRefGoogle Scholar
  35. 35.
    Godard, C., Roger-Estrade, J., Jayet, P. A., Brisson, N., & Le Bas, C. (2008). Use of available information at a European level to construct crop nitrogen response curves for the regions of the EU. Agricultural Systems, 97(1–2), 68–82.CrossRefGoogle Scholar
  36. 36.
    Llewelyn, R. V., & Featherstone, A. M. (1997). A comparison of crop production functions using simulated data for irrigated corn in Western Kansas. Agricultural Systems, 54(4), 521–538.CrossRefGoogle Scholar
  37. 37.
    Frank, M. D., Beattie, B. R., & Embletton, M. E. (1990). A comparison of alternative crop response models. American Journal of Agricultural Economics, 72(3), 597–603.CrossRefGoogle Scholar
  38. 38.
    Schott, C., Mignolet, C., & Benoît, M. (2009). Agriculture du bassin de la Seine. Programme PIREN-SEINE No.5, Agence de l' eau Seine-Normandie. http://www.sisyphe.upmc.fr/piren/fascicules. Accessed 28 July 2010.
  39. 39.
    Viennot, P., Ledoux, E., Monget, J. M., Schott, C., Gernier, C., & Beaudoin, N. (2009). La pollution du bassin de la Seine par les nitrates. Programme PIREN-SEINE No.3, Agence de l' eau Seine-Normandie. http://www.sisyphe.upmc.fr/piren/fascicules. Accessed 28 July 2010.
  40. 40.
    Berntsen, J., Petersen, B. M., Jacobsen, B. H., Olesen, J. E., & Hutchings, N. J. (2003). Evaluating nitrogen taxation scenarios using the dynamic whole farm simulation model FASSET. Agricultural Systems, 76(3), 817–839.CrossRefGoogle Scholar
  41. 41.
    Schou, J. S., Skop, E., & Jensen, J. D. (2000). Integrated agri-environmental modelling: a cost-effectiveness analysis of two nitrogen tax instruments in the Vejle Fjord watershed, Denmark. Journal of Environmental Management, 58(3), 199–212.CrossRefGoogle Scholar
  42. 42.
    Goetz, R. U., Schmid, H., & Lehmann, B. (2006). Determining the economic gains from regulation at the extensive and intensive margins. European Review of Agricultural Economics, 33(1), 1–30.CrossRefGoogle Scholar
  43. 43.
    Horner, G. L. (1975). Internalizing agricultural nitrogen pollution externalities: a case study. American Journal of Agricultural Economics, 57(1), 33–39.CrossRefGoogle Scholar

Copyright information

© Springer Science+Business Media B.V. 2012

Authors and Affiliations

  1. 1.UMR 210 INRA—AgroParisTech Economie PubliqueCentre de recherche de Versailles-GrignonThiverval-GrignonFrance
  2. 2.Department of Agricultural Economics and Rural DevelopmentAgricultural University of AthensAthensGreece

Personalised recommendations