AGORA Workshop Hammamet, Tunisia, October 2 nd, 2010 Developing models to estimate taxable revenues. The case of Pennsylvania forests Manuela Oliveira, University of Évora Marc McDill, Penn State School of Forest Resources Laura Leite, Penn State School of Forest Resources Jose G. Borges, Technical University of Lisbon CIMA - UE
Multidisciplinary expertise Manuela Oliveira Mathematics and Statistics http://www.uevora.pt Marc McDill Operations research and forest management planning http://sfr.psu.edu/directory/mem14 Laura Leite Growth and yield modeling Jose Borges - Operations research and forest management planning http://www.isa.utl.pt/home/node/369
Summary The context and our objective Pennsylvania forestry The need to estimate taxable revenues Materials and methods The forest inventory Database management data classification and aggregation Weighted linear regression Ongoing work
From Grace 2010
From Grace 2010
16.6 million ac.(6.7 million ha.) of forest (58% of the state) More forest today then in the last 150 years Maturing forest World-class quality hardwoods Suite of uses and values Vital to the economy and quality of life From Finley 2007
From Grace 2010
Nationally, Pennsylvania ranks 17th in terms of the area of forest land * However, in terms of hardwood growing-stock volume, Pennsylvania ranks #1 in the USA From Grace 2010
north and northwest central and southern areas From USDA Forest Service 2002
Forestland ownership 4% 23% Federal State 54% From Butler 2007 2% 17% Local Businesses Families and individuals (over 500,000 owners)
Total Income from Forest Land Over the Last 10 Years $200,000 or more 2% Missing 11% $100,000-$199,000 1% $50,000-$99,999 2% $10,000-$49,000 9% $1,000-$9,999 12% None 59% From McDill 2007 $1-$999 4%
The need to estimate taxable revenues Local governments and school districts depend on property taxes for revenue. The amount of property tax a landowner owes depends on the assessed value of their real property (land, buildings, and other improvements) and the local tax rate. Assessed values are normally based on the fair market value of each property. However, forest and farm landowners in Pennsylvania whose properties meet certain criteria can enroll in a preferential tax program commonly known as Clean and Green From Jacobsen and McDill 2008
Basic concept of Clean and Green Development Pressure From McDill 2007 Higher Land Values Higher Taxes Landowners are Forced to Develop Open Space Disappears
Participation in Clean and Green 2.7 million acres of forestland in 29 counties were enrolled in Clean and Green in 2004 55 counties have C&G The amount of tax savings depends on: The extent to which development pressure pushes up fair market values above use values The length of time since a re-assessment In many counties there is no benefit because the last reassessment was done long ago. In some counties, C&G values were once less than FMV but are now higher. From McDill 2007
How to estimate taxable revenues? Forest values depend in large part on the species composition of the stand. In order to account for this variation for property tax purposes, six forest types have been defined for Pennsylvania: Softwoods Miscellaneous Hardwoods Select Oaks (>50% Red) Other Oaks Northern Hardwoods Black Cherry (>40%) From Jacobsen and McDill 2008
Materials and methods The forest inventory and database FIA is a program of USDA Forest Service and is providing the most comprehensive forest ecosystem inventory in the US. Inventories since 1997 have used a common design for plot and tree data. Goal is to re-inventory each permanent plot every five years.
Materials and methods Database management data classification and aggregation Identification of 8,539 combinations Plot x Condition with Age > 0 our observations Identification of 64 forestypes and 136 forest species to be aggregated later into 6 forest type groups Identification of 287,967 tree records
Materials and methods Database management data classification and aggregation Query example SELECT [Cond_Manuela with foresttype descriptions].plt_cn, [Cond_Manuela with foresttype descriptions].plot, [Cond_Manuela with foresttype descriptions].condid, [Cond_Manuela with foresttype descriptions].condprop_unadj, [Cond_Manuela with foresttype descriptions].fortypcd, [Cond_Manuela with foresttype descriptions].meaning, [Cond_Manuela with foresttype descriptions].stdage, [Cond_Manuela with foresttype descriptions].sicond, [Cond_Manuela with foresttype descriptions].siteclcd, Tree_Manuela_Plot_Volume.PLOT_COND_VOLUME FROM [Cond_Manuela with foresttype descriptions], Tree_Manuela_Plot_Volume WHERE [Cond_Manuela with foresttype descriptions].plt_cn = Tree_Manuela_Plot_Volume.PLT_CN AND [Cond_Manuela with foresttype descriptions].condid = Tree_Manuela_Plot_Volume.CONDID;
Materials and methods Weighted linear regression There are advantages to accurate weight functions in linear regression: estimates of the variance-covariance matrix are biased if the error structure is incorrectly specified; a better fitting weight function leads to improved precision in estimation and is more likely to produce reliable confidence intervals.
Materials and methods Weighted linear regression Volume as a function of Age for each of the forest type groups An observation: one combination plot x condition Weight proportion of a condition in a plot (area weight) * age weight Plot x Condition Plot
Ongoing work Functions will be adjusted as data aggregations are completed e.g. IF Z = 501 THEN IF Percent_ Volume_N_Red_Oak >.5 OR Percent_Volume_White_Oak >.5 THEN ForestType_Group = 2 ELSE ForestType_Group = 3 END IF END IF
Ongoing work Prices will be multiplied by volume to obtain value functions Assessment of yield taxes as an alternative to property taxes
Thank you CIMA - UE