In the past, I’ve written several articles and posts on farmland markets in Oregon. All of these have relied on the results of surveys conducted by the U.S. Department of Agriculture, which ask producers to report their best estimate of the market value of their land. Naturally, this brings up the question of how well these survey responses capture the price at which land would sell in a market transaction. This post will briefly shed some light on that by comparing recent trends in USDA survey responses and observed, actual farmland transaction prices in Oregon.
The sales data I’m using come from CoreLogic, a private data vendor that assembles information on housing and land sales from each county in the US. Currently, the sales transactions cover the most recent transaction of all parcels labeled in the database as “agricultural” through July of 2021. In the next few months, I’ll have access to a larger database that covers land sales up through 2023 and includes both the most recent and historical sales of all farmland in the state, but for now this is what I have. To make an apples-to-apples comparison, since sales prices do not split out the price of any buildings or infrastructure present on the parcel, the survey data measure I’m using is the average farm real estate value, which represents the combined value of both land and farm-related buildings.
Because the database requires some cleaning, in this initial post, which I plan to be the start of a longer series of posts and articles on farmland prices in Oregon, I remove observations that might reasonably be considered outliers. First, I adjust the prices and survey values for inflation using the 2023 Gross Domestic Product Implicit Price Deflator. I then remove all sales with a per-acre price less than $100/acre or above $50,000/acre, as very low prices would potentially indicate sales between family members that are not at full market value and very high prices likely indicate sales that are not made for a bona fide agricultural use. I also remove sales that are less than 20 acres or above 5,000 acres. Although sales outside of these acreage bounds might reflect true agricultural sales, I do this to be conservative. Small-acreage sales often point to land in uses that are not truly agricultural in nature and large sales might reflect data processing errors.
To further minimize the presence of sales made for immediate non-agricultural use, using GIS data from the Department of Land Conservation and Development I remove sales of parcels inside of an urban growth boundary at the time of sale. For each year, I then take the average acreage-weighted sale price for all sales taking place in that year, which gives me the average price for an acre of land as opposed to the average unadjusted sale price.
The statewide trends in survey values and observed transaction prices are shown in the above figure. While both point to a general upward trajectory in farmland values over the 1997-2021 period, it is clear that survey values fall considerably short of average observed sales prices in each year. In a given year, sales prices are, on average, 78% higher than the survey value. Over the most recent five years (2017-2021), the average sale price amounts to $5,057/acre, while the average self-reported value from the USDA is $2,935/acre. The sales sample size ranges from 362 in 2009, during the depths of the Great Recession, to 821 in 2020, the most recent complete year available. I expect the sample sizes to be larger once I have the complete sales histories put together over the coming months. However, farmland markets tend to be fairly thin, with relatively few land parcels up for sale each year (e.g., compared to housing markets), which partly explains the more erratic trend in observed prices.
What might explain the large discrepancy between self-reported land values and actual sales transaction prices? For one, because farmland markets are thin, it could be a case of sales selection bias, where only higher-quality parcels are likely to be bought and sold. It could also be the case that sales data are skewed towards land bought by investors or other market participants who plan to develop land, which would lead to a price premium. Although I remove sales inside UGBs, all else constant, we would still expect the price of land close to existing UGBs to be bid up relative to land further away. Producers responding to surveys may also underestimate the market value of their land, especially if they are not fully considering the land’s potential in a future developed or other non-agricultural use. At any rate, this brief analysis makes clear that survey-reported land values and sales prices are capturing different features of Oregon’s land market.
Going forward, I plan to delve more deeply into the sales data to look at relationships between land prices and things like urban proximity, irrigation water rights, drought, and other ongoing policy issues in Oregon’s farmland market.