By Marie Coffin
It takes advanced modeling, complex regressions, and other high-level calculations to arrive at our valuation outcomes. Below are some of the methods we use — a general overview of approaches can be found here: Modeling the Determinants of Farmland Values in the United States (2013).
Hedonic Regression Approach
Hedonic models are widely used in real estate to estimate the selling price of an individual property based on certain predictive or explanatory regressors. (For example, sales prices for single-family dwellings might depend on number of rooms, number of bathrooms, lot size, school district, and distance to nearest grocery store.) These models are typically fit using some sort of regression, such as multiple linear regression, ridge regression, generalized linear regression, etc. As a result, we obtain the contribution of each explanatory variable to the overall price.
- A GIS-based Hedonic Price Model for Agricultural Land (2015)
- Fritz M. Roka, Raymond B. Palmquist; Examining the Use of National Databases in a Hedonic Analysis of Regional Farmland Values, American Journal of Agricultural Economics (1997)
- Environmental Amenities and Agricultural Land Values: A Hedonic Model Using Geographic Information Systems Data
- Valuation Using Hedonic Pricing Models (2009)
Tax Assessor’s Approach
The County Assessor’s Valuation method is used by many tax assessors to assess farmland values. It is based on three key pieces of information: the average value of farmland in the region (e.g. county), the average productivity rating for the region, and the weighted productivity rating of the individual parcel. It can be used with any productivity rating, such as CSR2 in Iowa or CPI in Minnesota, and the results are independent of the scale of the productivity rating.
In the first step, we calculate the average value of a productivity point:
Value Per Productivity Point = Avg Farmland Value in County / Avg CSR in County
In the second step, we apply this to the weighted productivity rating of the parcel:
Value = Value Per Productivity Point * Weighted productivity for Parcel
Some key features of this approach include:
- It can easily work with any productivity rating;
- Various annual survey datasets exist for county-wide average farmland values, across the country;
- The approach is sensitive to the macroeconomic factors at the county level; and
- Within a county, parcel value is purely a function of agronomic value.
Income Capitalization Approach
This method views farmland as an investment, and assesses its investment value based on the expected returns (i.e., the present value of a future cash flow).
The classic present value equations are:
When n → ∞ (i.e. cash flow runs to perpetuity), we have
Thus, we can estimate the present value of farmland by estimating the yearly cash flow and the rate of return. Because this is an investment approach, the rate of return can be estimated as the long-term rate of return on another investment of comparable risk (e.g, stock or bond).
This approach is well defined and has the appearance of mathematical validity. However, estimating the cash flow on a farm is a non-trivial problem. At CIBO, we have made some headway into obtaining the data required to estimate operating costs per parcel. The next step would be to estimate yearly yields and price fluctuations in order to estimate cash flow. Because this is a complicated and error-fraught problem, however, we have put this pursuit on hold for the moment.
Comparable Sales Approach
Comparables are widely used in the housing market to level-set asking prices. The general practice consists of locating similar houses (same or nearby neighborhood, same size, same age and condition, etc.) that have sold recently, and using those selling prices as a benchmark. In the housing market, this is often one of several inputs to the final asking price, but it is not unreasonable to arrive at a valuation by simply taking a weighted average of selling prices of comparables.
Applying this approach to farmland sales is less straightforward:
- Sparsity of data: In any given year, only a small number (less than 1%) of farms are bought and sold. Comparables are usually applied on a very local level (to account for macro- and micro-geographic trends). On a local level, the number of farm parcels that have sold recently is usually quite small. When one tries to further restrict this to comparable farm parcels, the data may become vanishingly sparse.
- Lack of definition: In residential and commercial real estate, the factors that contribute to “comparable” are well defined, generally agreed upon, and easily measured. There is no such general agreement with regard to agricultural land. There are a large number of factors that might make farms comparable, and some parcels are more easily measured than others. A first step to applying this approach is to perform a study of possible factors —this, in itself, is a form of hedonic regression (see above).
CIBO intends to explore the possibilities of the comparable-sales approach by building regression models to determine the factors of comparability.
Nearest Neighbor Interpolation
Nearest neighbor interpolation is related to the comparables approaches, and it operates by taking a weighted average of the n closest sales records as an estimate of the sale price of a parcel. Initial exploration of this method showed that the sparsity of sales data caused problematic results. Geographic trends dictate that this approach will only work if the n closest sales records are clustered closely around the parcel in question. Because sales data is so sparse, there will often be few records geographically close enough to be relevant, and the small sample size makes the interpolation behave erratically. For the moment, we have deprioritized this approach.
Historically, most approaches to estimating farmland value have been at the macroeconomic scale, looking at spatial-temporal trends in farm prices vs inflation factors. These approaches generally provide broad insights into boom-bust cycles in farmland value, but are not fine-grained enough to differentiate between one farm and another in the same county.
About Marie Coffin
Marie Coffin is the VP, Science and Modeling at CIBO, a science-driven software startup. She has focused on being a biostatistician at agriculture companies. Prior to CIBO, she worked for Monsanto, Icoria, Paradigm Genetics, and was an assistant professor at Clemson University. She holds a BS in Mathematics from South Dakota State University and a Ph.D. in Statistics from Iowa State University.