Mario Ortiz – Daniel Páez lab.uniandes.edu.co Context Estimated investment: $25 billion Purpose of 4G concessions is to improve national competitiveness…. …and extend national roads coverage Communications investment /GDP Transportation investment/GDP Energy investment/GDP National road network Secondary road network Tertiary road network Privates Source: National Infrastructure Agency (ANI) Fuente: DNP, Ministry of transport. 2011 Context Primary funding for the 4G road concessions would come from priovatizacion of ISAGEN, the largest power energy generator and owner of the distribution system of the country Estimated sell value: $2.5 billion Problem As opportunities for privatization are reduced, developing nations need to find new sources of funding for public infrastructure All land value increases are a consequence of national development and public sector investment Land value capture consist on obtaining a portion of the benefits cause by the new infrastructure on land prices Several examples of land value capture exist in urban areas There are opportunities to investigate further land value capture opportunities on rural areas Problem Before land values can be captured, a consistent, defendable and practical methodology is needed to determine land value increases due to the new infrastructure Infrastructure scenarios In this research proposes a methodology to estimate land value increments due to new national road infrastructure Determination of land value increases Develop land value capture mechanism Objetives Concessions National Roads ESTIMATED KILOMETERS INTERVENE ESTIMATED TOTAL INVESTMENT (BILLION / DIC - 11) To estimate economic impact of new national roads on land values Current Concessions Source: ANI New Concessions We focused on the new generation of national road concessons called 4G (forth generation) Methodology Design variables and determine model Valorization Analysis according to 4G Standard Linear Regression model (OLS) Applying the GWR Model 4G Simulation Spatial Analysis and Results Based on: Páez & Currie (2012) Previous experiences With a similar methodology we have estimated increases of land values for Bogota Modeling with GWR Geographically Weighted Regression (GWR) Ordinary Least Square (OLS) GWR provides a local model of the variable to analyze, adjusting the regression equation by entity. GWR uses the geographic principle that things that are near influence each other Model _ = 0 +1 _ + 2 _ + 3 _ + 4 (_ ) + 5 _ + 6 _ + 7 _ + 8 ℎ_ + 9 ℎ_ + 10 VARIABLE NAME SOURCE YEAR Dependent land values IGAC 2011 Explanatory Rural Aqueduct Coverage SSPD 2008 Explanatory Industrial and Commercial Income DNP 2009 Explanatory Rural Electric Coverage DANE 2005 Explanatory School Non-Attendance DANE 2005 Explanatory Rural Unsatisfied Basic Needs DANE 2010 Explanatory Subsided Health Regimen Coverage HealthMin 2010 Explanatory Rural Population DANE 2005 Explanatory Main Road Averaged Area of Influence OSM 2013 Explanatory Secondary Road Averaged Area of Influence OSM 2013 Explanatory Tertiary Road Averaged Area of Influence OSM 2013 Variables RURAL UNSATISFIED BASIC NEEDS (UBN) INDEX 100% Coverage (%) Rural power Coverage 80% 60% 40% Quindío Valle Caldas Guajira Guainía Vichada Chocó 20% 0% Departments Rural potable water Coverage Socioeconomic Land values Rural Unsatisfied Basic Needs (UBN) Index Rural power Coverage Rural potable water Coverage Land values Variables Influence area of the road network by municipalities Road Infrastructure Primary roads Secondary roads Tertiary roads Road network influence radius through a coverage area depending on the type of road (Primary, Secondary, and Tertiary). Average radius of 40 km per road Primary roads Secondary roads Primary roads Secondary roads Tertiary roads Tertiary roads Results Representation of the model in 45% of the real variation of the land value in a rural zone. OLS: Ordinary Least Squares 1 R² 45,6% Regression Statistics Multiple correlation coefficient (R) R Square Adjusted R Square Standard Error Observations ANOVA - Variance Analysis Degrees freedom (df) Square Sum (SS) Mean Square (MS) The problem with the model is related to the influence of the variables in the land value change in rural areas globally, it does not allow to analyze the inequality of land value in the whole territory. Consequently, the variables without statistical significance were removed, resulting in a new model. 0.680051286 0.462469751 0.456793612 1.006832351 958 Regression Residual Total 10 825,9328241 82,59328241 947 959,9846797 1,013711383 957 1785,917504 F Significance F 81,47613196 2,1907E-120 ANOVA - Variance Analysis Interception Primary roads Secondary roads Tertiary roads Industrial and Commercial Income Aqueduct Coverage Electric Coverage Rural Unsatisfied Basic Needs School Non-Attendance Health Coverage Rural population Coefficients Standar Error t Statistic 17,8152 0,1515 117,6078 Probability - 0,1291 0,0218 5,9319 0.0000 (0,0150) 0,0157 (0,9611) 0,3367 (0,0016) 0,0027 (0,6001) 0,5486 0,0318 0,0060 5,2776 0.0000 0,0024 0,0020 1,1948 0,2325 0,0027 0,0020 1,3167 0,1882 (0,0111) 0,0028 (4,0262) 0.0001 (0,0600) 0,0097 (6,1973) 0.0000 0,0052 0,0026 2,0452 0.0411 0,0001 0,0000 16,3817 0.0000 The model does not represent a significant difference with respect to the specification and statistical significance. OLS: Ordinary Least Squares 2 R² 45,5% Without the insignificant variables there is an explanation of 45% of the variation of land value in rural zones. The OLS model does not allow the use of variables that were considered important in the beginning (influence of secondary and tertiary roads, and the public utilities coverage). The analysis of these variables is important for a country with high geographical diversity, this being the reason they will taken into account in the GWR model. Regression Statistics Multiple correlation coefficient (R) R Square Adjusted R Square Standard Error Observations 0.6775 0.4580 0.4555 1.0080 958 ANOVA - Variance Analysis Degrees freedom (df) Square Sum (SS) Mean Square (MS) Regression Residual Total 6 819,6559 136,6093 951 966,2616 1,0160 957 1785,9175 Interception Primary roads (Line Density) Commercial Income (Log) Rural Unsatisfied Basic Needs School Non-Attendance Health Coverage Rural population F Significance F 134,4516 0.0000 Coefficients Standar Error t Statistic Probability 17,7669 0,1469 120,9616 0.0000 0,1159 0,0074 15,5587 0.0000 0,0329 0,0059 5,6197 0.0000 (0,0149) 0,0020 (7,4116) 0.0000 (0,0501) 0,0085 (5,9232) 0.0000 0,0093 0,0017 5,6030 0.0000 0,0001 0,0000 16,9095 0.0000 Geographically Weighted Regression R² 51,7% Features GWR Model Proyection 4G Local R2 The GWR model increases the explanation of the land value variable with respect to the independent variables (45% to 50%) The sum of the residual squares is lesser than the OLS model, it shows a lower level of error in the spatial analysis. Higher explanation of the land value and lower prediction error in the Andean and Pacific regions. On the contrary, the Caribbean and Orinoquía regions show a low level of explanation and a higher measure of error. Standard Error 4G Simulation Influence area of the road network by municipalities Road network and 4G concessions Analysis by department Land value and net valorization in rural areas comparison from the current road network and the inclusion of 4G projects Primary roads + 4G 4G roads Results Rural land appraisal after 4G simulation High net valorization of land in the central zone, where the majority of the projects are concentrated Departments with higher valorization: Cundinamarca, Valle del Cauca, Quindío and Risaralda. Additionally Bogota Distrito Capital. Boyacá, Risaralda and Nariño show the highest percentage valorization with respect to their original value (>2%) Cundinamarca and Valle show the highest valorization (approx. $10 million COP of additional value per km2) Highest Valorization Land value difference/Km2 Net appraisal (Billion COP) DEPARTMENT Nariño Chocó Chocó Nariño Nariño MUNICIPALITY Magüi Payan Río Iro Medio San Juan Providencia Gualmatán VALORIZATION 14,28% 12,32% 8,59% 8,40% 8,31% Conclusions The 4G concessions have a generally positive economic impact in the country, however the additional value is not equitable in the regions and is mainly concentrated in the central zone, gradually diminishing outward toward the eastern and northern periphery. The GWR model shows the main explanation of the land value and a lower prediction error in the Andean and Pacific regions. In contrast to that, in the Caribbean and Orinioquía regions it shows a lower power of explanation and a higher measure of error. The 4G road concessions have a high economic impact over the land valorization generally in rural areas. The GWR model allows to execute a more in depth analysis of the importance of each explanatory variables for every point in Colombia than the OLS model does. In a country with high geographical, cultural, demographical, and other types of diversities; the relevance of a variable can change radically from one place to another. GWR is a great tool to help make decisions in terms of infrastructure.