Distance to Biorefinery Plants and Its Influence on Agricultural Land Value: Evidence from the United States Midwest Region

Abstract

This paper examines the influence of distance to biorefinery plants on agricultural land value. The research utilizes micro-level data on characteristics of parcels and the locations of ethanol facilities in Central Nebraska, USA. I employ a theoretical model and an empirical hedonic approach to determine the effect of proximity. To address potential endogeneity, the econometric model adopts an instrumental variable. Estimates from the model provide evidence that land values significantly decrease as the distance from a parcel to a biorefinery plant increases. The results also reveal a non-linearity pattern in the model, with land prices falling at a decreasing rate with distance from ethanol plants.

1. Introduction

With the growth of the renewable fuels industry in the United States, and the increasing use of corn to produce ethanol in particular, questions have arisen regarding the impact on corn prices, farm profitability, and the value of farmland. This research examines agricultural land values and the impacts of proximity to ethanol plants. The study focuses on the Central Platte Natural Resources District (CPNRD) of Nebraska, a region with significant corn and ethanol production. The econometric approach estimates a hedonic land value model to determine the effect of proximity to ethanol plants, finding that property values are substantially affected by distance to this key local market for corn and other crops. The estimates account for the endogeneity of biorefinery plant locations using an instrumental variable and also permit non-linearity in the land value-distance relationship.

1.1. Relevant Literature

The context is concern regarding rapidly rising land values, with agricultural land values rising by 177.8 percent in the Corn Belt and 189.4 percent in the Northern Plains between 2007 and 2021, respectively [1]. Low interest rates and the adoption of productivity-enhancing technologies supported an increase in farmland values. Global factors of agricultural demand such as increases in population and income per capita also drove up agricultural commodity prices and land values [2]. Energy policies and tax credits for biofuels supported biofuel production, with benefits for land values, particularly in regions within higher local ethanol plant capacity [3,4,5,6]. Corn used for ethanol fuel production was 5.74 billion bushels, which outweighed the amount of corn directly used for feed and residual use (5.67 billion bushels) among the total amount of 16.3 billion bushels produced in 2021 [7].

1.2. Methods

In order to examine the influence of proximity to biorefinery plants on agricultural land value, the research adopts both theoretical and empirical modeling approaches. The hedonic model that consists of a function of the productivity characteristics of land is applied to the case of agricultural land values in the paper. Specifically, the crop price is determined by the distance from the agricultural and to the local market. Furthermore, a farmer can reduce transportation costs when his or her agricultural land is located close to biorefinery plants and other local processors of crops, and the land value can be captured with land characteristics and other regional economic factors.

The hedonic model estimates 2018 land values for agricultural parcels within a single region based on distance to the nearest ethanol plant. It also takes into account the endogeneity of biorefinery plant locations by employing an Instrumental-Variable (IV) approach utilized to study land use choices [8,9]. In addition, this research tests for non-linearity in the effect of distance on land prices based on the predictions of the theoretical model. Non-linearity implies that land values fall at a decreasing rate with distance from an ethanol plant. Further, to estimate the hedonic land price model, I utilize precise information on land quality and access to irrigation in parcels using Geographic Information System (GIS) databases. The GIS data provide a more cost-effective data source for such land characteristics than producer surveys. The results suggest that researchers and policy makers can adopt this approach with flexibility in choosing a study region, whether or not producer surveys have been taken.

1.3. Hypothesis

This research paper hypothesizes that agricultural land values decrease as the distance to a biorefinery plant increases. The transportation costs in hauling bulky crops and product are associated with the location of a biorefinery plant, and thus, proximity to agricultural markets increases farmland values due to savings in transport and hauling costs.

2. Background on Farmland Value and Its Determinants

Researchers have shown that the existence of ethanol plants has effects on land use and land value. However, early studies failed to account for the endogeneity of the location of ethanol plants. These studies found mixed results regarding whether proximity influences land values. A study shows that farmland within the Kansas City Federal Reserve District and finds higher land values during 2006 and 2007 for farms located in proximity to ethanol plants, particularly for agricultural lands located within 50 miles of an ethanol facility [4]. Another study, utilizing USDA Agricultural Resource Managerial Survey (ARMS), concludes that rental rates and the value of farmland close to ethanol plants are significantly higher than elsewhere [5]. Other study, using data on sales of 961 individual farmland parcels in northeastern Nebraska over the period 2004–2008, however, fail to find evidence that farmland values nearby biorefinery plants are higher than other comparable agricultural lands [10]. Similarly, research using county-level panel data in Iowa shows that there is no significant local effect on cash rental rates for cropland [11]. In addition, a study conducted in Wisconsin finds that biorefinery facilities have no significant effects on residential land values or agricultural land conversion in proximity to a plant [12].

More recent research accounts for the endogeneity of ethanol plant location within regions. For instance, when ethanol plants are located in regions with high corn productivity and associated high land values, research that finds proximity to ethanol plants is associated with higher land values may not be reliable. Therefore, it is essential to take the endogeneity of ethanol plant locations into account. An IV approach is needed to account for the possibility that ethanol facilities may be located in areas where land is more valuable because it is more productive for planting corn. For instance, a study finds higher land values for farms located within 30 miles of an ethanol plant than matched parcels not in proximity to an ethanol facility after the implementation of expanded Renewable Fuels Standards (RFS) [13]. They use a propensity score matching approach to address the endogeneity of plant location. Notably, they found no relationship between a regional ethanol plant and land values, potentially explaining inconsistency in results in the earlier literature.

Ethanol plant capacity also affects related factors such as producer decisions about land use. In the literature, a model of land use is built in the vicinity of ethanol plants where land use is a function of parcel characteristics and distance from the plant [14]. Other similar research study provides evidence that the share of land devoted to corn cultivation rises with regional ethanol plant capacity [15,16,17], although one of them found a smaller impact after accounting for the influence of statewide corn prices [15]. However, a recent study does not find a relationship between county ethanol plant capacity and re-enrollment rates of land in Conservation Reserve Program (CRP) [6].

Beyond region-wide impacts, some recent research has examined rent premium for proximity to ethanol plants within regions, that is, whether farmland values fall continuously with distance from a plant. Distance to a large local market might be expected to increase farmland values due to savings in hauling costs. Research finds corn prices increased by 12.5 cents per bushel nearby ethanol plants from the Corn Belt to the Upper Midwest [18]. Further, the coefficients for the impact of proximity on land value rise, as the definition of the impacted region is restricted to 10 miles from an ethanol plant rather than 20 or 30 miles [3]. Lastly, a recent study also analyzes the relationship between proximity to an ethanol plant and land use in Nebraska and found that the cultivation of corn drops in a non-linear manner with distance from an ethanol plant [9].

3. Modeling Approach: Foundations of Land Value and Rent

In the literature of housing and land prices, it is fundamental that the physical and spatial characteristics of the land determine its value. The term “hedonic prices” is introduced to understand how both consumers and producers make a decision in locations and land prices associated with the different factors of land parcels [19]. That is, the theory of hedonic prices can be formulated as an economic problem in which consumers maximize utility with a locational choice and producers maximize profit as they minimize the total production costs, which fulfills a duality relationship in Economics.

Researchers find that models of land value have been developed as a function of the productive characteristics of land in the literature [19,20]. I modify the hedonic model for the case of agricultural land values [20]. Farmland value depends on the proximity to local markets as well as the unique characteristics of farmland including the size of parcels, crop productivity and groundwater capacity [21]. To begin with, the model introduces a net price function of crops, which consists of the price of crop j and distance from farmland to market for farmer i:

$$p_{ij}\left(D_i\right)=Pric{e}_j-t{D}_i,$$

(1)

where $p_{ij}\left(D_i\right)$ represents the net price of the crop j for producer i, $Pric{e}_j$ is a regional commodity price and t$D_i$ represents producer i’s transport or hauling costs to the nearest market. Profit-maximizing producer i who considers his/her bid for rent faces the following problem,

$$\begin{array}{c}max\\ \mathit{y}\end{array}{\pi }_i={{\displaystyle \sum }}_j^n{p}_{ij}\left(D_i\right){\mathit{y}}_{ij},\text{subject}\mathrm{to}G(\mathit{y},\mathit{x})=0,$$

(2)

where ${\pi }_i$ is a “partial” profit function, which represents the difference between the value of outputs and the value of non-land inputs. The profit function is subject to the production function G given as G(y, x) = 0, where $\mathit{y}$ represents a vector of net outputs and $y_{ij}$ is an output when $y_{ij}$ > 0, otherwise $y_{ij}$ is an input. A vector of x represents farmland characteristics, which influence fertility and productivity of the agricultural land.

Accounting for the distance ($D_i$) from parcels of land to markets and transport costs per mile of output, profit-maximizing producer i solves the output supply and non-land input demand functions, ${\mathit{y}}^{*}$ = $\mathit{y}$(p, x), assuming that production skills are homogeneous among producers in the region. These can be substituted back into Equation (2), and it is rewritten as,

$${\pi }_i{}^{*}={{\displaystyle \sum }}_j^n{p}_{ij}\left(D_i\right)y_{ij}^{*}(\mathit{p}(D_i), \mathit{x})={\pi }_i{}^{*}(\mathit{p}(D_i), \mathit{x}),$$

(3)

where ${\pi }_i{}^{*}$(p($D_i$), x) is partial profits defined implicitly. One can now derive a “full” profit function, ${\pi }_i{}^{{*}^F}$, by subtracting the farmer’s rental costs for the land, R(x), from the implicit partial profit function,

$${\pi }_i{}^{{*}^F}={\pi }_i{}^{*}(\mathit{p}(D_i), \mathit{x})-R(\mathit{x}).$$

(4)

If the value of full profit function, ${\pi }_i{}^{{*}^F}$ is zero in a competitive agricultural market setting, then Equation (4) becomes,

$$R_i\left(\mathit{p}\right(D_i), \mathit{x})={\pi }_i{}^{*}\left(\mathit{p}\right(D_i), \mathit{x}),$$

(5)

where $R_i$(p($D_i$), x) is the farmer’s rental costs for the land and the term ${\pi }_i{}^{*}$(p($D_i$), x) is profit. Since parcels of land located at a greater distance to markets pay higher transport costs per mile of output to deliver, the rental costs decrease with distance.

Finally, the land value function, V, can be computed in perpetuity by dividing ${\pi }_i{}^{*}$(p($D_i$), x) by the discount rate r, yielding the expression:

$$V_i(\mathit{p}(D_i), \mathit{x}, r)=\frac{{\pi }_i{}^{*}\left(\mathit{p}\left(D_i\right),\mathit{x}\right)}{r}.$$

(6)

As profit maximization theory suggests, the profit function is non-decreasing and convex in output price, p($D_i$). Convexity occurs as producers face higher net prices, due to the proximity to ethanol plants, would invest more in the productive capacity of land and machinery used in farming. As previous research suggests, farmers located in regions surrounding ethanol plants respond with increased investments or other operational changes [3]. The first and second derivatives of the land value function are as follows:

$$\frac{d}{d\mathit{p}\left(D_i\right)}{V}_i(\mathit{p}\left(D_i\right), \mathit{x}, r)\ge 0$$

(7)

and

$$\frac{d}{d\mathit{p}\left(D_i\right)} (\frac{d{V}_i\left(\mathit{p}\left(D_i\right), \mathit{x}, r\right)}{d\mathit{p}\left(D_i\right)}) \ge 0.$$

(8)

Assuming that interest rates, input and output prices are common for all producers within a particular geography, one can empirically test the model using the following equation:

$$V_i={\beta }_0+{\beta }_1{D}_i+{\beta }_2{D}_i^2+{\beta }_3{C}_i+\gamma {\mathit{X}}_i+\delta {\mathit{Z}}_i+u_i,$$

(9)

where $V_i$ is the assessed land value per acre for producer i. $D_i$ is a variable that captures the distance of each land parcel to nearest ethanol plant, and ${\beta }_1$ measures its impact on land value and $D_i^2$ represents the square of distance to allow for non-linearity. Coefficient ${\beta }_2$ captures the non-linear effect of distance to the nearest ethanol plant. $C_i$ measures the distance of a land parcel to the nearest town. ${\mathit{X}}_i$ is a vector of individual parcel characteristics, and ${\mathit{Z}}_i$ is a vector of county characteristics, which controls for the influence on the land prices. $u_i$ is an unobservable error term. Equation (9) represents the hedonic model similarly in the literature by [4,10], but this research has modified it to include more detailed parcel characteristics and distance to the nearest ethanol facility.

The derivatives of the profit function predict the expected signs of estimated coefficients ${\beta }_1$ and ${\beta }_2$ in Equation (9):

$${\beta }_1 < 0 \mathrm{and} {\beta }_2 > 0,$$

(10)

where the sign of ${\beta }_1$ suggests that an increase in the distance between a farmland parcel and the nearest ethanol plant has a negative effect on the farmland values. That is, the land value decreases as distance to an ethanol plant rises. The sign of ${\beta }_2$ suggests that farmland values will fall at a declining rate, as the incentive to invest in machinery and farmland improvements drops as distance rises. In other words, land values are expected to have a semi-parabolic shape in the region surrounding an ethanol plant [9].

Although much research adopts the methodological approach in hedonic prices, alternatives for modeling the determination of locations and land prices can be found in the literature. For example, the present value of a perpetuity in land can be derived from an estimate of the value depending on its future productivity. Specifically, the present value of land in perpetuity is formulated as the returns to land in time divided by the discount rate or premium over time [18,20]. To test the hypothesis of proximity to biorefinery plants and its effect on agricultural land value, a hedonic price model is estimated using Equation (9) in this paper.

4. Study Area and Data

The study area is the Central Platte Natural Resources District (CPNRD) of Nebraska, USA. See the map of the CPNRD in Figure 1, a region with significant corn and ethanol production [22]. As one of 23 Nebraska Natural Resource Districts (NRDs), the CPNRD is the source of groundwater and wells from Nebraska’s major river, the Platte River. In Nebraska, these local governments are in charge of conservation, development, and utilization of agricultural and natural resources. There are 11 counties in CPNRD, and there are three ethanol plants [9]. I estimate the effect of proximity to local corn markets, based on ethanol plant locations, on land values in the CPNRD. Then, I compare the findings with the literature, which suggests that there is no impact of distance to ethanol plants on land values in Nebraska [10]. In addition, one study shows a similar result of higher land values for farms located within 30 miles of an ethanol plant [3].

To address how distance to a local market affects land values in the study area, the analysis relies on GIS databases, which are available to access micro-level information on the characteristics of farmland [9,17,23]. I utilize the characteristics of land use and land cover, the location of the parcel, parcel size, assessed land value, crop productivity and well capacity, respectively, in the analysis [24,25,26,27,28]. This approach is more objective and reliable than much of the previous research, which relies on self-reported information gathered through producer surveys [4,5,10].

The Nebraska Department of Revenue provides property values as assessed by county assessors in the state [26]. There are approximately 75,000 land parcels in the 11 counties in the CPNRD. Each county assessor evaluates land values uniformly and proportionately based on the characteristics of the land. With a restriction on the minimum size of parcels of at least 40 acres for the purpose of analysis of farmlands [10], this research includes approximately 13,700 individual land parcels in the study. Assessed property values are gathered for 2018.

The 2018 land use and land cover data are collected from the Geospatial Data Gateway (GDG) at USDA Natural Resources Conservation Service (NRCS) [24]. National Commodity Crop Productivity Index (NCCPI) are collected from USDA—Natural Resources Conservation Service to take account of a measure of crop productivity in each parcel for 2018, which is an important production input in farmland [6,13,17,29]. Since irrigation is also an important agricultural practice affecting land value, I collect information on groundwater from the Nebraska Department of Natural Resources to incorporate the well capacity that each parcel potentially maintains [28]. The well capacity variable directly identifies capacity in each parcel for 2018 rather than frequently used less precise measures such as the percent of irrigated acres in a parcel’s county [3].

In addition, county characteristics including population, per capita personal income, and the number of cattle are included in the analysis. Those variables control for the influence of economic factors on farmland value at the county level. Data for these controls are from the US Census, US Bureau of Economic Analysis, and USDA National Agricultural Statistics Service, respectively [30,31,32].

Using the straight-line distance method, I compute the distance from each land parcel to the nearest ethanol plant and use that as a measure of proximity [4,5,9,12,18]. The approach also adopts an additional control for distance to the nearest large town, taking into account bid-rent theory where proximity to the Central Business District (CBD) may affect agricultural land values [33,34,35,36]1.

Table 1. Descriptive statistics for variables of interest, 2018.

Variable	Mean	S.D.	Min	Max
Distance to Nearest Ethanol Plant (miles)	14.12	6.66	0.21	35.83
Distance to Nearest Large Town (miles)	19.86	14.12	0.1	56.66
Crop Productivity Index	40.69	10.18	5	80
Well Capacity (gallons per minute)	69.14	1212.69	0	133,088
Property (Parcel) Value (USD)	394,429.7	258,022.7	12,555	2,818,235
Parcel Size (acres)	138.04	94.04	40.01	754.68
Property Value per Acre (USD/acres)	3237.16	1648.61	221.21	9876.07
Population in County	30,507.29	21,116.39	2621	61,705
Per Capita Personal Income in County (USD)	41,490.11	5428.87	35,444	53,915
Number of Cattle in County (head, thousands)	129	84.14	27.5	285
Observations	13,719

Note: The data sources collected in the study analysis are as follows. Distance to Nearest Ethanol Plant and Distance to Nearest Large Town—GIS Workshop; LLC Land Use Development and Nebraska Ethanol Board [25,37], Crop Productivity Index—USDA Natural Resources Conservation Service [24], Well Capacity—Nebraska Department of Natural Resources [28], Property (Parcel) Value, Parcel Size, and Property Value per Acre—Nebraska Department of Revenue [26], Population in County—US Census [30], Per Capita Personal Income in County—US Bureau of Economic Analysis [31], Number of Cattle in County - USDA National Agricultural Statistics Service [32].

reports descriptive statistics for key variables of interest. The mean distance between farmland parcels and the nearest ethanol plant is 14.12 miles. The shortest distance to an ethanol plant is 0.21 miles, while the furthest distance is 35.83 miles in the sample. The control variable for distance to the nearest large town has an average distance of 19.86 miles and ranges from one-tenth of a mile to nearly 57 miles.

The crop productivity index for land parcel in the study ranges from 5 to 80 (on a scale that ranges from 0 to 100). A higher index value indicates better crop [17,29]. Since irrigation is an important input factor for production, I compute the well capacity of each parcel using information on static water levels, pumping rates, and well depth for each parcel. The mean well capacity is 69.14 gallons per minute. Notice, however, the extremely large variability in well capacity, which indicates that the input cost of irrigation varies greatly across land parcels.

Table 1 also indicates that the average land value is about USD 3237 per acre. This is in line with other evidence of average land values. According to the University of Nebraska—Lincoln Extension, the average value of dryland cropland (with irrigation potential) in Central Nebraska was USD 3120 per acre for land with irrigation potential [38]. Lastly, the county characteristic variables show that the mean population in the 11 counties is just over 30,500, and the average per capita income is approximately USD 41,490. The counties in the study area had an average of 129,000 head of cattle in 2018, and I use cattle density (the number of cattle divided by the population in county) in the models as a proxy measure to capture the effect of feedlots and livestock on land values.

5. Results

5.1. Baseline Results

Table 2. Hedonic estimates of effect of proximity to ethanol plants on farmland values (USD/acre).

Variable	(1) OLS Estimates	(2) OLS Estimates	(3) IV Estimates	(4) IV Estimates
Parcel characteristics
Distance to Nearest Ethanol Plant (miles)	−82.86 *** (7.1)	−127.74 *** (7.6)	−297.74 *** (54.6)	−379.32 ** (64.8)
$(\mathrm{Distance} \mathrm{to} \mathrm{Nearest} \mathrm{Ethanol} \mathrm{Plant}{)}^2$	0.12 (0.2)	2.84 *** (0.3)	6.68 *** (1.7)	10.79 * (2.1)
Distance to Nearest Large Town (miles)	−2.33 ** (0.9)	−2.35 *** (1.0)	−11.69 *** (2.6)	−10.89 ** (2.4)
Well Capacity (gallons per minute)	−0.85 *** (0.2)	−0.75 *** (0.2)	−0.69 *** (0.2)	−0.57 *** (0.2)
Crop Productivity	64.8 *** (1.2)	61.18 *** (1.2)	65.08 *** (1.2)	61.92 *** (1.3)
Well Capacity * Crop Productivity	0.02 *** (0.01)	0.02 *** (0.01)	0.02 *** (0.01)	0.02 *** (0.01)
County characteristics
Population in County	0.01 *** (0.005)		0.01 *** (0.001)
Per Capita Personal Income in County (USD)	−0.03 *** (0.003)		−0.01 (0.01)
Cattle Density	−2.93 *** (0.3)		−1.40 *** (0.5)
Intercept	2971.03 *** (167.8)	1827.23 *** (80.9)	3451.26 *** (210.9)	3780.97 *** (512.5)
County Fixed Effects	No	Yes	No	Yes
Sample Size	13,719	13,719	13,719	13,719
Weak ID F-test (Cragg–Donald Wald)	Not applicable	Not applicable	123.64	101.03
$\mathrm{Adj}. R^2$$(\mathrm{or} \mathrm{Centered} R^2$)	0.34	0.40	(0.30)	(0.35)

Notes: The table reports coefficients from hedonic estimates. Standard errors are in parentheses. *** Significant at the 1 percent level. ** Significant at the 5 percent level. * Significant at the 10 percent level.

presents hedonic estimates for the effect of proximity to ethanol plants on land values. Columns (1) and (2) present ordinary least squares (OLS) results, and columns (3) and (4) report instrumental variable (IV) results.

Model (1) includes county characteristics, but no county fixed effects, while model (2) includes county fixed effects, but no county characteristics. In both of these OLS models, the estimated effect of distance to the nearest ethanol plant is negative and significantly different from zero. Hence, the results provide evidence that land values fall with distance away from ethanol plants. The distance squared term is not significantly different from zero in model (1), indicating a linear patter, although the squared term is positive and significant in model (2), indicating a non-linear parabolic spatial pattern. Distance to the nearest large town is negative and significantly different from zero in both models, indicating that the value of land parcels declines with distance from city centers, as expected from bid-rent theory. Well capacity and crop productivity are clearly related as indicated by the significance on each of these variables and their interaction. Crop productivity increases land value, as does the combination of crop productivity and well capacity. The county control variables in model (1) indicate that parcels in more populous counties are more valuable, as would be expected. While I would expect per capita personal income to also have a positive effect, its coefficient estimate is negative. Note, however, that the effect of this variable disappears in the IV estimates, discussed below. Finally, the cattle density control variable has an expected negative effect, as grazing land generally is less valuable than cropland. This result may also be expected due to the externalities involved. The literature suggests that there may exist a negative externality associated with the area with biorefinery plants. For instance, the federal and state laws require environmental considerations for the location of feedlots including degradation of water quality as well as odor and dust [39]. Furthermore, with the clusters of economic activities including hauling bulky agricultural product and transportations, it creates a negative externality and air pollutions.

Models (3) and (4) are the IV estimates. Parcel distance to the nearest railroad track is utilized as an IV for distance to the nearest ethanol plant, following the literature that utilizes railroad capacity in a region as an instrument for ethanol plant capacity [17]. This approach addresses potential endogeneity bias with OLS estimates in models (1) and (2)2. Since the model uses both distance to ethanol plant and distance-squared variables, I need two instrumental variables with two first-stage equations to address endogeneity problems in such models. Using the predicted values for distance to an ethanol plant and its squared term, $\widehat{D_i}$ and ${\widehat{D}}_i^2,$ respectively, I estimate the following equation:

$${Value}_i={\beta }_0+{\beta }_1{\hat{D}}_l+{\beta }_2{\hat{D}}_i^2+{\beta }_3{C}_i+\gamma {\mathit{X}}_i+\delta {\mathit{Z}}_i+u_i,$$

(11)

where $Valu{e}_i$ represents farmland values (USD/acre) for parcel i. The coefficients ${\beta }_1$ and ${\beta }_2$ capture the impact of distance and distance squared to the nearest ethanol plant on land values, respectively. The variable $C_i$ measures the distance of a given land parcel to the nearest city (large town). ${\mathit{X}}_i$ is a vector of characteristics of individual land parcels and ${\mathit{Z}}_i$ is a vector of county characteristics. The unobservable error term is denoted $u_i$.

I further conduct Cragg–Donald (CD) Wald F-statistic tests for weak instruments in the IV approach. The test results are reported in columns (3) and (4) of Table 2, in which the instrumental variables, Railroad and $Railroa{d}^2,$ are employed to control for the possible endogeneity bias with the locations of ethanol plants. The CD Wald F-statistic tests report the null hypothesis of a zero coefficient for the IV. A recognized general rule is that if the value of the F-statistic exceeds 10, the instrumental variable is considered valid and sufficiently strong to address the endogeneity issue. As reported in columns (3) and (4) of Table 2 and

Table 3. Tests for underidentification and weak instruments.

Variable	Underidentification		Weak Identification
	Sanderson-Windmeijer (SW) Chi-sq (1) p val		SW F (1, 13,709)
Distance to Nearest Ethanol Plant (miles)	248.25	0.0000	248.07
${\left(\mathrm{Distance} \mathrm{to} \mathrm{Nearest} \mathrm{Ethanol} \mathrm{Plant}\right)}^2$	253.48	0.0000	253.30
Stock–Yogo weak ID F test critical values for single endogenous regressor:
	10% maximal IV size		19.93
	20% maximal IV size		8.75
Underidentification test
Ho: matrix of reduced form coefficients has rank = K − 1 (underidentified)
Ha: matrix has rank = K (identified)
Anderson canon. corr. LM statistic	Chi-sq (1) = 243.08		p val = 0.0000
Weak identification test
Ho: equation is weakly identified
Cragg–Donald Wald F statistic			123.64

Notes: The underidentification test result indicates that I can reject the null hypothesis that the matrix of reduced form coefficients has rank = K − 1 (which means the equation estimation is underidentified). The week identification test result for IV indicates that I can reject the null hypothesis that equation is weakly identified. I conduct an additional test for underidentification, and the result (presented in Table 4) indicates that we can reject the null hypothesis that the matrix of reduced form coefficients has rank = K − 1 (or we can reject the null hypothesis that the equation estimation is underidentified).

, the F-test results confirm that the railroad and $railroa{d}^2$ instrumental variables employed in the study are valid3.

Focusing on the IV results in models (3) and (4), the estimated coefficients indicate that the influence of distance to the nearest ethanol plant on farmland values is negative, significant, and nonlinear. Both coefficient estimates are statistically significant in both versions of the IV models, with the distance coefficient negative and the distance squared coefficient positive. At the mean distance of 14.12 miles from an ethanol plant (from Table 1), using the coefficient estimates from model (3), I compute that an increase in distance of one mile reduces land value by about USD 109 per acre. This result is consistent with the previous research, which finds that the rental rates and farmland values close to ethanol plants are significantly higher [4,5]. Note also that the influence of distance to an ethanol plant on land values is much larger in the IV specification in column (3) relative to the OLS results in column (1). In the OLS results, given the coefficients for distance to nearest ethanol plant and its squared term, an increase in distance of one mile from an ethanol plant reduces land values by approximately USD 79 per acre at the mean distance of 14.12 miles. Consequently, the IV estimates indicate that the effect of proximity to nearest ethanol plant at the mean distance is about 1.37 times larger than the OLS estimate with no IV is employed.

The coefficient of crop productivity has a positive effect on land values, indicating that more productive land is more valuable. The coefficient on the interaction between well capacity and crop productivity is also positively related to farmland value. This finding is consistent with the theory that crop productivity and well capacity are important complementary input factors which farmers consider.

The coefficients for the county characteristic variables indicate that increased population has a positive effect on land values, as in the OLS model (1). This finding is expected because greater population shifts the demand curve for land and thereby affects the market value of land. An increase in personal income per capita has an unexpected negative, but an insignificant effect on land value. An increase in cattle density in the county is negatively associated with farmland value as in the OLS model (1).

A second model using the IV estimation approach is reported in column (4), which excludes variables for county-level characteristics but utilizes county fixed effects. In this case, I find a similar impact of distance to the nearest ethanol plant on farmland values. The coefficients of variables of interest show that farmland values are affected by both the distance and distance-squared to the nearest ethanol plant. At the mean distance of 14.12 miles, for instance, an increase of one mile further away from the nearest ethanol plant has the effect of decreasing land value by USD 74.61 per acre. The estimated coefficients for other explanatory variables are similar to those in column (3). For example, the coefficient on the interaction between well capacity and crop productivity continues to be positively correlated with farmland value.

In summary, the empirical evidence from the OLS and IV hedonic models (1)–(4) reported in Table 2 is consistent with the theory that property values in the region are substantially affected by proximity to local markets such as ethanol plants. The estimates indicate that land values fall with distance to the nearest ethanol plant and do so at a decreasing rate. The evidence on the negative relationship between distance to nearest ethanol plant and land values with a semi-parabolic shape is presented in Figure 2.

The figure shows the estimated land value at each distance using the results for model (3) from Table 2 and at mean values for all of the other variables. The non-linear relationship indicates the influence of distance on land values moderates as distance rises. For example, a parcel of agricultural land located one mile from the ethanol plant has the average value of USD 3160.2 per acre, holding other land characteristics constant at their mean values. Land value decreases to USD 2129.56 at 5 miles and to USD 1141.86 at 10 miles. Note that the estimated parabolic shape ultimately starts to rise, but this occurs well beyond the mean distance of 14.12 miles.

Figure 2 also compares non-linear IV results with those of IV estimates assuming a strictly linear relationship between proximity to an ethanol plant and land values. Allowing for a non-linear relationship between distance to the nearest ethanol plant and land values yields a substantially different spatial pattern in estimated land values. The non-linear results are consistent with previous findings in the literature [3]. Those authors examined the impact of ethanol plants on land values in regions surrounding a plant. Their robustness analysis found a larger increase in land values when the impact region is confined to a smaller 10-mile radius then when a 20- or 30-mile radius is used. Land parcel values within a 10-mile radius were 28 percent higher than values for parcels further from the ethanol facility. The point estimate dropped to 16 and 15 percent, respectively, when the impact region was expanded to a 20- and 30-mile radius.

5.2. Robustness Check

In this section, two robustness checks are conducted. First, the dataset as suggested in Table 1 indicates the existence of notable outliers. For example, one of the parcel observations has the maximum property value of USD 9876.07 per acre. For this particular parcel, the total property value is USD 397,808, and the size of the parcel is 40.28 acres. To investigate whether such outliers affect the estimation results, I compute distributions of the observations with kernel density estimation to detect outliers (observations that have a standardized z-score greater than 3 or less than −3). Figure 3 illustrates the kernel density estimation. There is evidence of 56 observations that have a standardized z-scores greater than 3 in the sample. Hence, I conduct a robustness check by omitting those 56 outliers, re-estimating the IV models with results reported in

Table 4. Robustness check: IV estimates of the effect of proximity to ethanol plants on farmland values (USD/acre).

Variable	(1) IV Estimates	(2) IV Estimates	(3) IV Estimates	(4) IV Estimates
Parcel characteristics
Distance to Nearest Ethanol Plant (miles)	−310.22 *** (53.4)	−394.51 ** (63.43)	−293.36 *** (51.98)	−395.48 *** (63.49)
$(\mathrm{Distance} \mathrm{to} \mathrm{Nearest} \mathrm{Ethanol} \mathrm{Plant}{)}^2$	7.06 *** (1.6)	11.26 * (2.01)	6.95 *** (1.56)	11.26 *** (2.00)
Distance to Nearest Large Town (miles)	−12.17 *** (2.5)	−11.29 ** (2.38)	−10.80 *** (2.42)	−11.16 *** (2.36)
Well Capacity (gallons per minute)	−0.71 *** (0.2)	−0.58 *** (0.22)	−0.75 *** (0.22)	−0.59 *** (0.21)
Crop Productivity	64.74 *** (1.2)	61.92 *** (1.23)	60.52 *** (1.18)	61.61 *** (1.21)
Well Capacity * Crop Productivity	0.02 *** (0.01)	0.02 *** (0.01)	0.02 *** (0.006)	0.02 *** (0.006)
Ethanol Capacity			4.65 *** (0.24)	−5.94 *** (1.27)
County characteristics
Population in County	0.01 *** (0.0007)		0.02 *** (0.0007)
Per Capita Personal Income in County (USD)	−0.003 (0.01)		−0.02 *** (0.006)
Cattle Density	−1.24 ** (0.5)		−1.04 ** (0.49)
Intercept	3367.34 *** (207.0)	3902.29 *** (502.24)	3562.87 *** (203.10)	4287.34 *** (556.02)
County Fixed Effects	No	Yes	No	Yes
Sample Size	13,663	13,663	13,663	13,663
Weak ID F-test (Cragg–Wald)	123.67	100.93	123.63	101.19
$\mathrm{Adj}. R^2$$(\mathrm{or} \mathrm{Centered} R^2$)	(0.30)	(0.35)	(0.33)	(0.35)

Notes: The table reports coefficients from hedonic estimates. Standard errors are in parentheses. *** Significant at the 1 percent level. ** Significant at the 5 percent level. * Significant at the 10 percent level.

and comparing the results to those in Table 2. The new outcomes consistently report that land prices fall at a decreasing rate with distance from ethanol plants. The results for the robustness check confirm that the findings are not influenced by the existence of outliers.

The second robustness check is related to a concern over heterogeneity in ethanol plant capacity. It may be the case that the effect of proximity to ethanol plants is affected by the capacity of those plants. To investigate whether the results are sensitive to such heterogeneity, I further collect data on the capacity of each ethanol facility, and the analysis includes a variable for plant capacity of the nearest ethanol plant in the regression [37]. Land values would be expected to rise with ethanol plant capacity since larger facilities would offer a larger price premium to attract corn from a larger surrounding region. Including this variable in the model, the result indicates that its estimated coefficient is positive and statistically significant in the model controlling for individual county characteristics but changed signs in the model using county fixed effects. However, inclusion of the ethanol plant capacity variable in the model does not yield a significant change in the magnitude or significance of the coefficients on distance to the nearest ethanol plant or distance squared.

6. Discussion and Conclusions

This study examines whether proximity to local markets influences agricultural land value. I utilize micro-level data on characteristics of parcels and the sites of ethanol facilities in Central Nebraska for 2018. A hedonic model of land values is estimated to determine the effect of proximity to biorefinery plants. To address the possible endogeneity issue related to biorefinery plant location, the model utilizes an IV approach. The research also conducts robustness checks testing whether the results are influenced by the presence of outliers. The findings show evidence that land values in the region are substantially affected by distance to this key local market. Land values fall with increasing distance to the nearest biorefinery plant and do so at a deceasing rate.

Previous research has shown mixed results in the relationship between proximity to a biorefinery plant and the value of surrounding land [4,5,10,11,12,18]. Employing GIS databases on the characteristics of parcels, I develop a hedonic model with the locations of biorefinery plants [9,17]. Estimates from the hedonic model indicate that land values in the study region significantly decrease as the distance from parcel to markets increases. As a result, the importance of the transportation costs associated with different locations is reflected in property values in the study area, where agriculture is a major industry. In Europe, especially with some empirical cases of Italy, researchers also discuss the importance of the provision of renewable energy and seek for the benefits of the bio-based economy as well as the impact of agricultural policy on energy efficiency and sustainability [40,41,42,43]. Similar to the findings from this research, a non-linear effect on rental prices of farmland is found from Northern Italy [44].

This paper contributes to the literature in several ways. First, I utilize IV methods in a model of land value and distance to market, especially biorefinery plants. Previous models using IV methods assumed a constant impact of biorefinery plants on land values within a defined surrounding region, with the size of the impact varying with biorefinery plant capacity. Second, consistent with theory, the results suggest a non-linear relationship between distance to a biorefinery plant and land values, and I find that the spatial distribution of estimated land values varies significantly between linear and non-linear models. Third, by using GIS databases, this research is able to incorporate other hedonic characteristics such as well capacity and crop productivity in the hedonic models. This unique approach has the potential to be applied in other studies of the influence of biofuel production on land values within agricultural regions.

The proximity to biorefinery plants and its impact on agricultural land value provides the broader application and policy implications. Specifically, a close distance to a biorefinery plant increases not only the price of the agricultural land but the portion of land cultivated for corn and other crops such as soybeans in the US Midwest Region [17]. In the agricultural area where biorefinery plants are dominant for the local development, it tends to increase the production for corn and other crops as well as the allocation in agricultural land within the parcels close to the biorefinery plants, but reduction in cultivation for grassland and pasture resulted from the increasing land value [9].