Zoppi & Lai - Input2012


Published on

Corrado Zoppi and Sabrina Lai on "Empirical evidence on agricultural land-use change in Sardinia (Italy) from GIS-based analysis and a Tobit model"

Published in: Technology, Business
1 Comment
  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

Zoppi & Lai - Input2012

  1. 1. Seventh International Conference on Informatics and Urban and Regional Planning Cagliari, May 10-12, 2012 Planning support tools: policy analysis, implementation and evaluation Empirical evidence on agricultural land-use change in Sardinia (Italy)from GIS-based analysis and a Tobit model Corrado Zoppi & Sabrina Lai Università di Cagliari - Dipartimento di Ingegneria Civile, Ambientale e Architettura Via Marengo, 2 – 09123 Cagliari, Italy Tel.: +39 070 6755216/6755206, telefax: +39 070 6755215 zoppi@unica.it; sabrinalai@unica.it
  2. 2. Layout1. Introduction2. Methodology: Tobit model3. A GIS-based taxonomy of Sardinian municipalities4. The impact of investment in public services and infrastructure from the 2000-2006 ROP-EAGGF5. Discussion and conclusions: Tobit models are a useful tool to analyze and interpret spatial phenomena in conjunction with GIS representations
  3. 3. 1.a – Introduction • to assess whether, and to what extent, the 2000-2006 ROP-EAGGF was effective in maintaining agricultural land use compared to other physical, economic and social characteristics concerning local development. Aim • a Tobit model combined with Geographic Information System.Method • the towns of the Sardinian Region.Case-Casestudy • some interesting insights concerning investment policies to maintain, reinforce and possibly expand agriculture land use, whose relevance can be clarified only through a spatial representation. • the model’s results show evidence of a strong correlation between the stability/increase of agriculturalResults land use and the ROP-EAGGF investment, even though its impact is quantitatively less important with respect to other physical, economic and social characteristics concerning local development.
  4. 4. 1.b – Introduction: the 2000-2006 ROP-EAGGF. The period to assess land-use changeA fundamental pillar of the 2000-2006 cohesion policy of the EU in Sardinia,the 2000-2006 ROP was a plurifund program based on: • EAGGF – European Agricultural Guidance and Guarantee Fund (approx. 770 million euros, 20 % of the total investment of the program) • ERDF – European Regional Development Fund • ESF – European Social Fund • FIFG – Financial Instrument for Fisheries GuidanceMore than 90% of the expenditure of the 2000-2006 ROP-EAGGF occurredin the period 2003-2008: • measures aimed at simplifying payment procedures were adopted only in 2003: only from that year did the program really start • expenditure of the 2000-2006 ROP had to be entirely realized by Jun 30, 2009 due to the impossibility to meet the original deadline of Dec 31, 2006 2003-2008 is the most suitable time period to assess if the 2000-2006 ROP-EAGGF succeeded in maintaining agricultural land use
  5. 5. 1.c – Introduction: the 2000-2006 ROP-EAGGF investment Measures of the 2000-2006 ROP-EAGG (Regione Autonoma della Sardegna, 2009a, pp. 300, 313, 367-391) Expenditure Measure Type of operations (million €)1.2 - Integrated cycle of water resource 43.494 Public infrastructure and servicesmanagement: irrigation systems of agricultural zones1.9 - Prevention and control concerning forest fires, 11.000 Public infrastructure and servicesand reforestation4.9 - Investment in agricultural firms 183.092 Funds granted under the aid regime rules4.10 - Improvement of transformation and marketing 137.848 Funds granted under the aid regime rulesof agricultural products4.11 - Marketing of agricultural products 27.796 Funds granted under the aid regime rulescharacterized by a high qualitative level4.12 - Diversification of agricultural activities and the 10.000 Funds granted under the aid regime ruleslike4.13 - Essential services for the economy and the 15.659 Public infrastructure and servicesrural population4.14 - Promotion of adaptation and development of 31.740 Public infrastructure and servicesrural areas4.17 - Restoration of the rural environment damaged 16.000 Public infrastructure and servicesby fire and natural disasters, and prevention4.18 - Professional education referred to all of the 6.000 Education2000-2006 ROP-EAGGF measures4.19 - Rural reparcelling 59.957 Feasibility studies. Funds granted under the aid regime rules4.20 – Development and improvement of rural 143.626 Public infrastructure and servicesinfrastructure for agriculture4.21 - Young farmers’ start-up support 84.325 Funds granted under the aid regime rules
  6. 6. 1.c (continued) – Introduction: the 2000-2006ROP-EAGGF investmentThe thirteen measures of the 2000-2006 ROP-EAGGF show acomprehensive approach which entails addressing severalaspects of rural development: conservation of natural resources and of rural cultural heritage protection from fire and natural disasters restoration and renovation of rural buildings and settlements capacity building and professional education tourism and tourism-related activities reforestation support to agriculture, in terms of infrastructure and services
  7. 7. 1.c (continued) – Introduction: the 2000-2006ROP-EAGGF investmentObjective not to evaluate the comprehensive impact and effectiveness of the program but only to assess its impact on agricultural land useTherefore, only the subset of measures that finance publicinfrastructure and services for agricultural land use is hereconsidered, that is measures 1.2, 1.9, 4.13, 4.14, 4.17, 4.20.Assumption: the other measures (funds granted to agriculturalfirms under the aid regime rules, and to promote professionaleducation) do not influence agricultural land use, at least in theshort run, even though they could possibly have an impact onrural economy.The investment expenditure referred to the six measures isabout one third of the total investment.
  8. 8. 2.a – Methodology: a Tobit model The spatial configuration of agricultural land-use change between 2003 and 2008 in Sardinian towns is considered to be dependent on the following variables:Variable Definition Mean St.dev.PLUAA3_8 Percentage change in a municipalitys agricultural land use 4.6660 17.2699 between 2003 and 2008DENS Residential density of a municipality in 2008 (residents per 0.7744 2.0981 hectare)VARUU3_8 Percentage change in a municipalitys urbanized land 22.8301 41.1650COAST Dummy - A municipality overlaps a coastal landscape unit 0.4430 0.4974 as defined by the Regional Landscape PlanPERVARPI Percentage change in a municipalitys per-capita income in 21.8675 9.5568 the period 2003-2008EXPEAGF 2000-2006 ROP-EAGGF per-capita expenditure for 456.1207 669.9753 measures 1.2, 1.9, 4.13, 4.14, 4.17, 4.20 concerning a municipality (thousand Euros) The most relevant methodological point of this study is to demonstrate how the Tobit model can be used to address an important problem of spatial analysis.
  9. 9. 2.b – Applications of Tobit modelsconcerning rural development Gebremedhin and Swinton (2003): tenure rights and farmers attitude towards soil conservation, Northern Ethiopia Alene et al. (2000): relationship between adoption and intensity of utilization of improved maize varieties, and household characteristics, West Shoa, Central Highlands, Ethiopia Coomes et al. (2000): relationships between non-market mediated access to land and labor, and forest fallow management and duration, Peruvian Amazon Pfaffermayr et al. (1991): assessment of the labor-supply decisions of farm workers between part-time off-farm and full-time in-farm options, Austria Baidu Forson (1999): application of a Tobit model to analyze the determinants of level and intensity of adoption of innovative technologies for soil conservation and water resource management, Niger Rajasekharan and Veeraputhran (2002): relationship between intercropping decisions and availability of family labor and the type of intercrops during the initial gestation periods of natural rubber cultivation, Kerala Region, India The Tobit model used in this paper is based on Greene (1993, pp. 694-700)
  10. 10. 2.c – The model We consider a censored or latent dependent variable, y*, •y=L if y* ≤ L which is related to an observable variable, y, as • y = y* if y* > L (1) follows: The model operationalizes by assuming that y* is linearly • y* = β’x + ε (2) dependent on a vector of explanatory variables, • where β = (β1,…, βm) is a x = (x1,…, xm), through the vector of coefficients following relation:
  11. 11. 2.c (continued) – The modelIf we substitute (2) in (1) we obtain the following model, which, under the usualhypotheses of ordinary least-squares models concerning the mean, variance andcovariance of the ε term, can be estimated by the maximum likelihoodestimator (MLE), which is unbiased and efficient, even though non-linear(Greene, ibid.): y=L if y* ≤ L y = β’x + ε if y* > L (3)The MLE for β is the vector b = (b1,…, bm) that maximizes the following maximumlikelihood function: (4)where s2 is the MLE estimator for the variance of the error term and Φ is thecumulative normal distribution operator (Greene, ibid.).The values of the vectors of coefficients bj’s which maximize (4) are the solution ofthe system which comes from equalizing to zero the derivatives of ln L with respectto the m components of vector b.
  12. 12. 2.c (continued) – The modelThe values of the estimates bj’s of the vector of coefficients βj’s make itpossible to estimate the marginal effects of a change of the vector ofexplanatory variables x on the censored variable y as follows (Greene, ibid.): (5)The model assumes that: the error term ε is normally distributed E(ε|x) = 0 (i.e., the error terms in the regression have a 0 conditional mean) Var(ε) = σ2 (i.e., the error term has the same variance at each observation) E[ εi εj | X ] = 0 (i.e., the error terms are uncorrelated between observations) s2 = where y is the vector of observations of the censored variable y and x is the matrix (n,m) of the n observations of the m explanatory variables.
  13. 13. 3.a – A GIS-based taxonomy ofagricultural land use & related variablesAim: to analyze each city/town changes in land use byimplementing a spatial databaseBasic spatial units: municipalitiesIntegration of available (both spatialand non-spatial) data, e.g. spreadsheets (demography, income, levels of expenditure, …) Need spatial datasets (distribution of land uses, for delimitation of coastal landscape units, …)was required to develop new knowledge GISand obtain new layers of eitherspatial or non-spatial information
  14. 14. 3.b – Land cover and land-cover changes1990-2006: data produced by theEuropean Environmental AgencyFreely downloadable fromwww.eea.europa.euDistinction between real evolutionprocesses and differentinterpretations of the samesubjectNot detailed enough (scale,minimum mapping units) 2003 & 2008 regional Land cover changes involving areas classed as agricultural between 1990 and 2000 and between 2000 and 2006 “land-use maps” according to EEA’s land-cover maps
  15. 15. 3.c – 2003 and 2008 regional land-usemaps Freely downloadable from http://www.sardegnageoportale.it 4 levels hierarchical nomenclature, compliant with the CORINE Land Cover project strictly speaking, the so-called “regional land-use maps” are actually land-cover maps … … however, thanks to their level of detail, they do also provide reliable information on how land is used by humans, especially as far as agricultural areas and forests are concerned Land-use changes in the 2003-2008 time the expression “land-use map” is frame, taking into account only areas that in 2003 were classed as agricultural according therefore used here (but we are to the regional land-use maps aware of the difference!)
  16. 16. 3.d – Agricultural land uses in Sardinia according to the 2003 and 2008 land-use maps Total area Total area 2nd level 3rd level 4th level in 2003 in 2008 [hectares] [hectares] 2111 Non-irrigated arable land 144,537.8 251,181.45 211 Non-irrigated arable land 2112 Artificial meadows 142,587.2 164,575.47 2121 Arable land and21 346,524.70 205,735.63 horticultural crops in open fieldsArable land 212 Permanently irrigated land 2122 Rice fields 4,584.17 4,660.69 2123 Nurseries 240.08 341.00 2124 Crops in greenhouses 1,184.98 1,769.0422 221 Vineyards 15,957.87 24,686.43Permanent 222 Fruit trees and berry plantations 10,268.18 11,907.84crops 223 Olive groves 43,790.91 48,777.6223 Pastures 231 Pastures 9,517.21 10,316.08 2411 Annual crops associated 9,608.73 11,714.50 with vineyards 241 Annual crops associated with 2412 Annual crops associated 163.70 296.1324 permanent crops with vineyardsHeterogeneous 2413 Annual crops associated 53,164.42 58,067.58agricultural with other permanent cropsareas 242 Complex cultivation patterns 43,107.79 42,206.07 243 Land principally occupied by agriculture, with significant 27,271.17 29,282.01 areas of natural vegetation 244 Agro-forestry areas 50,493.25 57,429.67 Total 903,002.15 922,947.21
  17. 17. 3.e – Descriptive attributesISTAT census code of each municipality in the Italian Census system OriginalAREA_CITY land areaPERIMETER length of the boundaryNAME_CITY name of the municipalityPOP_2003 Resident population as of December 31, 2003POP_2008 Resident population as of December 31, 2008 Used in MNLDENS Residential density in each municipality in 2008 [residents/hectare] DerivedCOAST Dummy - A municipality overlaps a coastal landscape unit as defined by the RLPPLUAA3_8 Percentage change in a municipalitys agricultural land use bw 2003 & 2008VARUU3_8 Percentage change in a municipalitys urbanized land bw 2003 & 2008PERVARPI Percentage change in a municipalitys per-capita income in the period 2003-2008EXPEAGF 2000-2006 ROP-EAGGF per-capita expenditure for measures 1.2, 1.9, 4.13, 4.14, 4.17, 4.20 concerning a municipality (Euros per resident)VARPOP Percentage change in a municipality’s population bw 2003 & 2008AGR03 Total amount of agricultural land in 2003 [ha]PERCAGR03 Percentage of agricultural land in a given municipality in 2003AGR08 Total amount of agricultural land in 2008 [ha]PERCAGR08 Percentage of agricultural land in a given municipality in 2008LUURB_03 Total amount of urbanized land in 2003 [ha] Derived – not used in MNLPERCURB03 Percentage of urbanized land in a given municipality in 2003LUURB_08 Total amount of urbanized land in 2008 [ha]PERCURB08 Percentage of urbanized land in 2008LUC_AU Land-use change from “agricultural” to “urbanized” bw 2003 & 2008 [ha]LUC_AA Land-use change from “agricultural” to “agricultural” (≠sub-types) bw 2003 & 2008 [ha]LUC_AF Land-use change from “agricultural” to “forestry” bw 2003 & 2008 [ha]LUC_AS Land-use change from “agricultural” to “scrubs” bw 2003 & 2008 [ha]LUC_AB Land-use change from “agricultural” to “bare soil” bw 2003 & 2008 [ha]LUC_AP Land-use change from “agricultural” to “wetlands” bw 2003 & 2008 [ha]LUC_AW Land-use change from “agricultural” to “inner & maritime waters” bw 2003 & 2008 [ha]% LUC_AU Land-use change from “agricultural” to “urbanized” bw 2003 & 2008 [%]% LUC_AF Land-use change from “agricultural” to “forestry” bw 2003 & 2008 [%]% LUC_AS Land-use change from “agricultural” to “scrubs” bw 2003 & 2008 [%]% LUC_AB Land-use change from “agricultural” to “bare soil” bw 2003 & 2008 [%]% LUC_AP Land-use change from “agricultural” to “wetlands” bw 2003 & 2008 [%]% LUC_AW Land-use change from “agricultural” to “inner & maritime waters” bw 2003 & 2008 [%]TOTINC2003 Per-capita income in 2003TOTINC2008 Per-capita income in 2008
  18. 18. 3.e (continued) – Descriptive attributes: examples1. PLUAA3_8: % change in a municipality’s agricultural land use 2003-20082. DENS: residential density in 20083. VARUU3_8: % change in artificial surfaces 2003-2008
  19. 19. 3.e (continued) – Descriptive attributes: examples1. COAST: municipalities that overlap/do not overlap any coastal landscape units2. PERVARPI: % change in per-capita income 2003-20083. EXPEAGF: 2000-2006 ROP-EAGGF per-capita investment on public infrastructure and services for agriculture
  20. 20. 4. – The impact of investment in public servicesand infrastructure from the 2000-2006 ROP-EAGGF Hypothesis test: Variable Coefficient bi Standard error z-statistic βi=0 DENS 0.3479 0.0680 5.116 0.0000 VARUU3_8 0.0394 0.0220 1.791 0.0732 COAST 4.1603 1.8525 2.246 0.0247 PERVARPI 0.2073 0.0737 2.812 0.0049 EXPEAGF 0.0070 0.0007 10.434 0.0000 Marginal Hypothesis test: Variable Standard error z-statistic effect βi=0 DENS 0.3470 0.0678 5.119 0.0000 VARUU3_8 0.0393 0.0219 1.791 0.0732 COAST 4.1498 1.8466 2.247 0.0246 PERVARPI 0.2067 0.0736 2.809 0.0050 EXPEAGF 0.0070 0.0007 10.429 0.0000
  21. 21. 4. (continued) – The impact of investment in publicservices and infrastructure from the 2000-2006ROP-EAGGF• An increase of about 3.5% in • This variable has a positive and agricultural land would occur if significant marginal effect, which DENS increased by 10 residents puts in evidence that investment in per hectare, which puts in infrastructure and services to evidence a significant improve agricultural land use is agglomeration effect. more effective in coastal areas than in inner ones. The results imply, ceteris paribus, a 4% positive differential in agricultural land use.Residential A municipality overlaps a coastaldensity landscape unit (RLP)
  22. 22. 4. (continued) – The impact of investment in publicservices and infrastructure from the 2000-2006ROP-EAGGF• The change in a municipality’s • This implies not only that there is an urbanization also reveals a positive income effect that does not displace marginal effect, even though not at all agricultural activities (0.2% per quantitatively strong: approximately a 1% increase in RPCI), but also that, only a 3.9% of a percentage point for in principle, whichever policy aiming a 1% increase. However, this is at increasing local people’s income important since it shows that can be considered an indirect increasing urbanization does not support to maintaining and possibly seem to prevent conservation or increasing agricultural land use. expansion of agricultural land use.Change in a Real per capitamunicipalitys income increaseurbanization
  23. 23. 4. (continued) – The impact of investment in publicservices and infrastructure from the 2000-2006ROP-EAGGFFinally, the marginal impact of the investment concerning a municipalityfrom the 2000-2006 ROP-EAGGF is positive, as expected, and highlysignificant.However, this impact is quantitatively weak, since it indicates that a100€ increase in per-capita investment implies approximately a 0.7%increase in agricultural land use.In other words, this means that an investment of nearly 165 millionEuros, that is more than one fifth of the total 2000-2006 ROP-EAGGF expenditure, would have been required to increase agriculturalland in 2008 with respect to 2003 by approximately 63 km2.
  24. 24. 5. – ConclusionsThis paper proposes a discussion on quantitative change in agricultural land useat the municipal level as a phenomenon influenced by physical, economic, socialand investment policy variables. The methodological approach is based on theintermixed use of GIS techniques and econometric models.We feel that the use of a Tobit model based on a GIS-based taxonomy ofdependent and explanatory variables could be easily and effectively exported toother European Union regional contexts, since the 2000-2006 ROP’s of theEuropean regions at the NUTS 2 level were fairly standard.The reproducibility of the proposed approach makes it possible to assess theresults concerning the impact of ROP-EAGGF’s on agricultural land-use changeand to compare such impacts across regions, at the intra-national and inter-national levels.Moreover, the results are useful in terms of ex-post assessment definition and implementation of regional policies concerning investment aimed at maintaining and increasing agricultural land use, that is in terms of ex-ante and on-going assessment.
  25. 25. 5.(continued) – Conclusions: policy implicationsfrom the model resultsIn the rest of this concluding remarks we use GIS to comment anddiscuss policy implications of our results through a spatial representation.Background policy implication: it should be more effective to invest inagriculture in municipalities having significant values of residential density whose territory overlaps a coastal landscape unit“What-if” scenario built upon marginal effects from the Tobit model.Two steps:1. If a single explanatory variable increased by a given quantity (ten percentiles in its distribution) … … what would the magnitude of the impact on the % change in a municipality’s agricultural land use between 2003 and 2008 be?2. What impacts would be produced by implementing policies that increase four variables (DENS, VARUU3_8, PERVARPI, EXPEAGF)?
  26. 26. 5.(continued) – Conclusions: a spatial representation of policy implicationsImpacts on agricultural land use stemming from policies that increase…1. … a municipality’s residential density (Imp. DENS)2. … % change in a municipality’s urbanized land between 2003 and 2008 (Imp. VARUU3_8)3. … per-capita income at the municipal level (Imp. PERVARPI)4. … per-head investment on public infrastructure and services for agriculture (Imp. EXPEAGF)
  27. 27. 5.(continued) – Conclusions: a spatialrepresentation of policy implicationsThe map shows impacts on preservation ofagricultural land uses produced by implementingpolicies that increase: residential density urbanized land per-capita income per-capita investment on public infrastructure and services for agricultureIt therefore gives clear indications on which arethe “best” possible areas that policies should betargeted in order to preserve and reinforceagricultural land uses.