A spatial econometric approach to assess the impact of RDPs agri-environmental measures on the use of Nitrogen in agriculture: the case study of Emilia- Romagna (Italy) Marconi V., Raggi M. and Viaggi D. AIEAA conference, Parma,6-7 June 2013 Motivation and Objectives SPARD project - Spatial Econometric Analysis: RDPs evaluation Ex-post evaluation extension to Ex-Ante assessment This study focuses on the Agri Environmental Schemes of RDPs 2007-2013 in an Italian case study (Emilia-Romagna Region): assess if specific actions of measure 214 are effective in reducing the nitrate surplus in agriculture. provide a quantitative evaluation of the impact of the agroenvironmental measures on the reduction of agricultural inputs through spatial regression model. INORGANIC MODEL and ORGANIC MODEL
Background (AES) 214/1 (integrated farming), only in preferential areas. 214/2 (organic farming: restrains the use of nitrogen based fertilizers), 214/8 (extensive meadows: conversion of intensive crops into pasture, fodder crops and grasslands), 214/9 (protection of natural, semi-natural and agricultural landscape: restrains the extension of intensive cultivation), 214/10 (set aside of arable crops for environmental purpose: reduces the extension of intensive cultivation). The applications for the remaining sub-measures were given absolute priority if belonging to a preferential area: 1.NVZs 2.LFA 3.PROTECTED AREAS (NATURA 2000, AND LOCAL LEVEL) All the regular applications for measure 214 were funded by the RDP 2007-2013. Background (Emilia-Romagna) 2 fertilizers sale in Italy, together with the other region of the Po plain owns one third of the national livestock NVZs
Background (Gross Nitrogen Balance) Calculation of the gross nitrogen balance (OECD-EUROSTAT, 2007) Indicator(Inorganic Fertilizers) INPUT 1. FERTILIZERS First estimate of inorganic fertiliser use for the region [kg N]: Σ (crop(i) area in the region [ha] X Application rate of inorganic fertiliser for crop(i) [kg N/ha]) NOTES Adjustment factor: Province scale inorganic fertiliser sale / Σ (inorganic fertiliser use per municipality) The contribution to the balance of urban compost and sewage sludge is considered to be small and not having effect on the final balance Fertlizers do not include livestock manure (only inorganic)
Indicator(Organic Nitrogen) INPUT 2. LIVESTOCK MANURE Within each region for each livestock category: Estimated quantity of N in livestock manure for category(i) [kg N/year] = Number of animals of category(i) for the region [heads] X manure coefficient of category(i) [kg N/head/year] For each region: Estimated quantity of N in livestock manure [kg N/year] = Σ (estimated quantity of N in livestock manure per livestock category) [kg N/year] NOTES Some deduction are made for volatilisation of NH3 during storage Use experimental regional data for N excretion CRPA Methodology Data collection and analysis NUTS 4 Agricultural census 2000 and 2010 Uptake of AES schemes, RDPs 2007-2013 CMEF impact indicator (water quality): Gross Nitrogen Balance (baseline indicator 20) Inorganic fertilizers input Organic nitrogen input Spatial analysis: Variations between years 2000-2010 Descriptive statistics Global Moran s I (Moran, 1950) Local Indicator of Spatial Association (LISA; Anselin, 1995) Autocorrelation, clustering, heterogeneity(outliers) Spatial regression model (Le Sage, 1999) OLS SPATIAL LAG SPATIAL ERROR Indicators variations as dependent variables Uptake of submeasures 214/1,2,8,9,10 RDPs (2007-2013) as explanatory variables
Results: D Inorganic N, D Organic N (2010-2000) Results: spatial analysis DN INORGANIC DN ORGANIC queen 1 All the painted municipality are significant at least at 5% queen 2 queen 3
Results: regression models Variables have been tested for multicollinearity: Variance Inflaction Factor < 5 F test is significant only for INORGANIC MODEL (not including livestock density) Spatial coefficients are highly significant General fit improves in the spatial models Results: Inorganic Model
Results: Organic Model) Synthesis of results Both inorganic (-%) and organic (-9 %). nitrogen deriving from agricultural systems has overall decreased from year 2000 to 2010. Positive spatial autocorrelation of the variable D N Inorganic. Spatial models are more effective in explain the variability of the difference in inorganic nitrogen input between 2000 and 2010 with respect to OLS regression. Cold-spot clustering (Ferrara) overlaps the area where integrated farming (action 1) was most contracted. Hot spots clustering in the western provinces Uptake of actions 1 and 10 is significant in all the regression models. Spatial distribution of action 214-10 and the occurrence of factors not captured by model (e.g. slippage, spatially correlated errors) should be further investigate. Other significant variables: ratio of certified organic land on the UAA in year 2000 (-), Y NVZs (), population density (), farms < 5 ha (). Reduction of organic nitrogen links directly to the decrease of the livestock density in the considered time shift If LUs / UAA is not included in the model, the test F do not validate the model.
Conclusions 1. screening study on the spatial dimensions of RDPs effects, amenable of further improvements. Access to farm level information and genuinely measured impact parameters are key needs ensure a better exploitation the potential of spatial econometrics. 2. the spatial dimension of the dependent variable cannot be interpreted in a straightforward way as a result of spillover effects in the variable itself, but most likely as an effect of spillovers in its determinants or similar production conditions in neighboring areas (districts) 3. the dependent variables are not observed data but are rather derived through calculation, which can bring some bias due to the hypotheses adopted in the process. Thank you!!!
FARMER S CHARACTERISTICS POLICY VARIABLES STRUCTURAL VARIABLES INDIVIDUAL COMPANY / TOT COMPANIES BUDGET MEASURE 214 ( ) / UAA (ha) D UUA / TAA UNIVERSITY DEGREE 214 1 EXTENSION / UUA D AVERAGE FARM SIZE (ha) HIGH SCHOOL DIPLOMA 214 2 EXTENSION / UUA D NUMBER OF FARMS AGE BETWEEN 40 54 214 8 EXTENSION / UUA D LIVESTOCK / TOT FARMS 214 9 EXTENSION / UUA D LIVESTOCK UNITS / LIVESTOCK 214 10 EXTENSION / UUA D LIVESTOCK UNIT / UUA (LSU / ha) Y NVZ MUNICIPALITIES Y LFA MUNICIPALITIES D ORG. LAND / TAA D ARABLE / UAA D ORCHARDS/ UAA UUA BETWEEN 5 30 (HA) UUA LOWER THAN 5 (HA) N SURPLUS (kg) / UAA (ha)_2000 POPULATION DENSITY Y HILL Y MOUNTAIN
DIAGNOSTICS F statistics : significant OLS NORMALITY OF ERRORS: Jacques Bera (significant) HETEROSKEDASTICITY Random coefficients are are highly significant (Breusch pagan, ols Koenker bassett) Ols Robust test (White) N/A, ok SPATIAL DEPENDENCE (LIKELIHOOD RATIO TEST significant)