Retiring Land to Save Water: Participation in Colorado’s Republican River Conservation Reserve Enhancement Program

1. Introduction

Groundwater rights have been overappropriated in many parts of the western United States, which has led to concerns about future groundwater availability and surface water depletions to the extent that some states have experienced difficulty complying with interstate streamflow compacts. The U.S. Department of Agriculture (USDA), individual states, and local water conservation districts have increasingly utilized agricultural land retirement programs in an effort to conserve groundwater resources, reduce challenges with the conjunctive use of water resources, and meet streamflow compliance. These programs typically pay producers to convert irrigated land to a conservation practice and temporarily or permanently cease irrigation.

Despite a surge of resources devoted to addressing water conservation through agricultural land retirement, no research that we are aware of has evaluated participation in or the performance of these programs. The lack of rigorous studies makes it very difficult to assess the economic trade-offs associated with such programs. This paper addresses this gap by using well-level data to explore the factors that influence agricultural producer participation in the Conservation Reserve Enhancement Program (CREP) in the Republican River Basin of Colorado, a program designed specifically to reduce agricultural groundwater use. The data allow us to consider the impact of conservation incentive amounts in conjunction with both land and aquifer characteristics. Our analysis is an important step toward developing an understanding of how participation in a land retirement program designed to reduce water use depends on the physical features of the landscape and the implications this may have for program performance.

Currently, 33 states have approved CREP agreements in place, with the programs in four states (Colorado, Nebraska, Kansas, and Idaho) focusing primarily on retiring irrigated agricultural land with the objective of reducing water use. Highlighting the need for economic research on program performance, none of the CREP programs in the four states focusing on water use have achieved their acreage enrollment goals after approximately 10 years. In response, the incentive payments offered for participation have been increased in several states in an effort to encourage additional participation.

In addition to being the first study to evaluate land retirement aimed at reducing water use, our research improves on previous literature by using individual well-level data on enrollment. The use of parcel-specific data allows us to better evaluate the revealed microeconomic and geophysical underpinnings of enrollment behavior compared to previous studies, which have relied on aggregate enrollment rates (e.g., Konyar and Osborn 1990; Plantinga, Alig, and Cheng 2001; and Suter, Poe, and Bills 2008) or hypothetical survey responses (e.g., Kingsbury and Boggess 1999; Armstrong et al. 2011; Yeboah, Lupi, and Kaplowitz 2015). Also, when a program objective is to conserve water through land retirement, it is important to consider how aquifer characteristics impact enrollment (and therefore water conservation), because the volume of water used can vary substantially as aquifer characteristics change, whether for strategic reasons or because of how the characteristics affect well capacity and pumping costs (Pfeiffer and Lin 2012, 2014). In the area that we study in Colorado, incentive rates for program participation vary as a function of distance from the Republican River. This variation in incentive rates allows us to appropriately identify the effect of incentives on actual enrollment in a way that has not been feasible in previous observational or survey research.

The incentives for participating in the program are offered to agricultural landowners who have a permit to withdraw groundwater in the Republican River basin of Colorado. The participation models that we estimate utilize spatially explicit data on the aquifer and soil characteristics at each of the groundwater wells that are eligible for the program, in addition to the differing financial incentives offered at each well site. Based on the model results, we, not surprisingly, find that the financial incentives that are offered have a significant impact on enrollment rates in the program. In particular, we predict enrollment rates to increase by 0.087% with a $10 increase in the net present value of the incentives offered. Estimates like this could be used by program administrators to calibrate financial incentives to achieve targeted enrollment rates. Another important finding for policy makers interested in retiring water rights to conserve water is that aquifer characteristics, such as the level of saturated thickness, and soil variables, such as the land capability class, also significantly impact enrollment. In particular, wells with lower saturated thickness that irrigate less-productive land are more likely to be enrolled in the program. This finding suggests that program enrollment is more heavily distributed in areas with the lowest opportunity costs and that the social efficiency of the program depends on the relative social benefits associated with retiring wells with specific hydrologic and agronomic features.

2. Review of Relevant Literature

Our study is informed by previous economic research that has sought to explain landowners’ enrollment decisions in agricultural land retirement programs, as well as studies that evaluate the impact of land retirement on the provision of ecosystem services. Economic studies of participation in land retirement programs utilize several methods for understanding variation in enrollment rates, including surveys of hypothetical participation, simulations of landowner behavior, and observed county-level enrollment data. While the existing literature has identified factors that influence enrollment in land conservation programs, it has not focused on how enrollment interacts with spatially explicit measures of aquifer or soil characteristics.

Surveys of hypothetical participation ask producers to state whether they would enroll land in a given land retirement program for a specified annual payment. These surveys also typically gather information on farm and producer characteristics and, in some cases, environmental and conservation values. Kingsbury and Boggess (1999) find that increasing incentive payments leads to increases in the probability of hypothetical participation in one Oregon county but, curiously, reduces it in another. In addition, the availability of cost-sharing incentives, perception of an environmental problem, and placing greater importance on environmental quality increase the stated likelihood of participation. Research by Lynch and Lovell (2003) combines survey results with geospatial data to assess the likelihood of participation in a farmland preservation program in Maryland and finds a greater likelihood of participation among larger farms and farms that have a higher probability that future generations will continue farming. Survey results reported by Armstrong et al. (2011) find that producer attitudes toward the opportunity cost of enrolling land and resentment toward water management policies and institutions significantly impact hypothetical enrollment in the New York City CREP, which primarily seeks to improve water quality. Yeboah, Lupi, and Kaplowitz (2015) survey agricultural landowners in Michigan’s Saginaw River watershed about hypothetical participation in CREP and find that higher annual payments and shorter contract lengths increase the probability of enrolling, but one-time signing payments do not have a significant impact on the decision to enroll.

Numerical simulation has also been used to evaluate the factors that affect participation in agricultural land conservation programs. Lynch and Brown (2000) use an optimization model to simulate a representative agricultural producer’s decision to use land for production, sell for development, or enroll in the Maryland Chesapeake Bay CREP program. The authors then model the magnitude of acreage enrollment and the landowner’s decision related to the specific conservation practices. A simulation model is also used by Feng et al. (2006) to evaluate how a landowner’s decision to enroll in land retirement rather than a working lands conservation program influences the optimal allocation of conservation funds and the provision of environmental benefits.

Several studies use county-level enrollment data to estimate models of participation in land retirement programs. Konyar and Osborn (1990) find that Conservation Reserve Program (CRP) participation increases as payment incentives relative to the average net returns to agricultural production in a county increase. In addition, participation is found to be negatively related to average land value, farm size, and producer age in a county. CRP enrollment in metropolitan and nonmetropolitan Northeastern counties is evaluated by Parks and Schorr (1997). They find that counties with greater maximum allowable rental rates, crop production costs, proportion of low quality agricultural land, and proportion of idle land have greater enrollment rates in nonmetropolitan counties. In metropolitan counties, an increase in production costs or low quality land is estimated to increase land retirement, while higher average land values decrease county enrollment rates. Plantinga, Alig, and Cheng (2001) estimate supply functions for agricultural conservation land for nine major U.S. regions. For most regions, higher incentive payments, eligible land area, and average land capability class lead to significantly higher rates of enrollment. Research by Isik and Yang (2004) uses county-level CRP enrollment rates in 100 Illinois counties to illustrate the importance of option values in driving land retirement decisions. They show that the likelihood of CRP enrollment is reduced by increases in the uncertainty and irreversibility of CRP participation decisions. County-level CREP enrollment data are used by Suter, Poe, and Bills (2008); they construct the average annual payment offered to eligible landowners in a given county by weighting soil rental rates by the distribution of soil types in the area eligible for enrollment. The authors find that annual and one-time payment incentives both increase enrollment, while a variable measuring urban influence is negatively related to enrollment.

In this study, we use spatially explicit microdata to determine the location of each well that is eligible for the program and the value of the financial incentives offered at each well. It is the first study that we are aware of to use microdata to model enrollment in an agricultural land retirement program, which enables a more detailed understanding of how participation varies as a function of spatially explicit physical characteristics of the landscape. Additionally, our study is the first that we are aware of to analyze the performance of a land retirement program that specifically targets water conservation. Several previous economic studies have evaluated the design and efficacy of land retirement programs targeting improvements in water quality (e.g., Ribaudo, Osborn, and Konyar 1994; Khanna et al. 2003; Yang et al. 2003; Uchida, Xu, and Rozelle 2005; Luo et al. 2006), wildlife conservation (e.g., Van Buskirk and Willi 2004), and general ecosystem services (e.g., Feng et al. 2006; Johnson et al. 2012), but none have evaluated land retirement from the perspective of reducing water use. The factors motivating participation in and the performance of land retirement initiatives targeting water conservation are likely to be different from those that affect programs targeting improvements in water quality, biodiversity, and ecosystem services. This study therefore contributes to the agricultural land conservation literature by directly addressing the factors influencing participation in a program that focuses on water conservation.

3. Background

Colorado’s Republican River CREP is a state-federal program initiated in 2006 that is designed to remove land from agricultural production to improve environmental conditions and reduce water use. The stated objective of the program is to enroll 35,000 acres of irrigated cropland in specific counties within the Republican River Water Conservation District, as illustrated in Figure 1. Enrolled land must be converted from agricultural production to a conservation practice for the duration of the 15-year contract and cannot be irrigated.¹ Importantly, participants must cancel the water allocation associated with the retired land in perpetuity (USDA-FSA 2011). The implication is that once the contract has expired, the enrolled acreage can return to dryland but not irrigated production. This is an important distinction for Colorado’s Republican River CREP. Other CREP programs allow landowners to commence irrigation upon contract expiration or, as in the case of Colorado’s Rio Grande CREP, offer higher incentives for producers who voluntarily choose to permanently retire their water rights (USDA 2012).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Area Eligible for Colorado (CO) Republican River Conservation Reserve Enhancement Program (CREP)

Irrigation in the program area is derived almost exclusively from groundwater wells (groundwater accounts for more than 96% of all water rights). As such, in the analysis that follows the unit of observation is an individual well. Landowners with eligible wells are offered annual and one-time payments for participation. An important feature of the incentive payments for our research is that they vary based on the distance from the North or South Fork of the Republican River. The intuition for the structure of the spatial variation in the incentives is the program’s stated interest in addressing conjunctive use concerns. Wells closest to the river have the most direct impact on surface water flows and therefore provide opportunities to more rapidly address compliance concerns with the Republican River Compact between the states of Colorado, Nebraska, and Kansas, as well as surface water quality concerns. The exogenous nature of the spatial variation in the incentive payments provides an opportunity to appropriately identify the impact of the incentives on participation in the program, controlling for aquifer and soil characteristics.

The offered incentive payments vary by distance from the North and South Fork of the Republican River as a result of additional funds provided by the state of Colorado and the Republican River Water Conservation District. The annual payment that is offered for participation is $140 per acre for irrigated land less than one mile, $130 per acre for land between one and two miles, $120 per acre for land between two and four miles, and $115 per acre for land more than four miles from the North and South Forks. The one-time payments include a signing and installation incentive payment, as well as a water right retirement payment. The signing and installation incentive varies from $65 per acre for land less than one mile from the North and South Fork to $15 per acre for land greater than four miles. The water right retirement incentive payment, paid in the 5th, 10th, and 15th years of the contract, varies from $400 per acre for land less than one mile to $100 per acre for land more than four miles from the river. Figure 2 shows the well locations in the study area, as well as the buffers around the North and South Forks of the Republican River that determine the annual incentives. Table 1 provides a breakdown of the incentive payments, as well as a count of the eligible wells and the participation rate.²

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Eligible and Participating Well Locations in the Republican River Conservation Reserve Enhancement Program

4. Model of Participation

In this section, we describe the theoretical model that motivates the empirical analysis of program participation. We then present the empirical specification of the participation model.

5. Results

The coefficient estimates from the probit regression described in equation [4] are displayed in Table 3, which uses data from all eligible wells, and Table 4, which uses data only from wells within eight miles of the Republican River. To assess the robustness of our estimates, we provide coefficient estimates from three separate models. In particular, model (1) includes only the annualized incentive and saturated thickness variables. Model (2) includes all of the variables capturing incentive and aquifer characteristics, while model (3) includes these variables as well as the soil characteristics. By restricting the sample, Table 4 allows us to assess the extent to which the model results are robust to focusing only on the areas in close proximity to the two forks of the Republican River. One cost of this geographic restriction, however, is that it reduces our sample size by nearly 80% (from 3,673 to 854 eligible wells).

Across all of the model specifications, the coefficient on the net present value of the incentive payments variable is positive, statistically significant from zero, and relatively stable. The coefficient estimate of 0.00181 from the full model specification (Table 3, column 3) that uses the entire sample translates to an average marginal effect of 0.000087. To put this value in perspective, in the spring of 2016 the USDA announced a plan to increase the annual incentive payments offered by $65 per acre. This change will increase the net present value of the incentives offered by more than $700 per acre. If we assume that the marginal effect remains constant for this relatively large increase, we predict that the higher incentive payment will increase the enrollment rate by more than 6%, or roughly triple enrollment in the program. The indicator for whether a well is near a river or stream is also found to have a positive and significant impact on enrollment. This variable is included to account for wells that might be eligible for the cost-share incentives offered in riparian areas. The variable measures eligible riparian areas beyond just the North and South Fork of the Republican River. Nevertheless, in the restricted model, there are fewer wells that are eligible for the cost-share incentives but do not also receive higher annual payments. Therefore, some of the reduction in the magnitude of the coefficient on the NPV variable in the models with the restricted sample may be attributed to its correlation with the variable indicating proximity to rivers and streams.

In addition to the incentives offered, participation is also found to depend on the aquifer characteristics at a particular well location. In the models that utilize the entire sample of data, both the level of saturated thickness and the change in saturated thickness are found to have a significant influence on enrollment. The marginal effect of saturated thickness implies that a well with a saturated thickness that is 10% (13.4 feet) below the mean is predicted to have a rate of enrollment that is 0.38% higher. In other words, wells with less groundwater availability are significantly more likely to enroll in the program. Consistent with this, wells that have experienced declines in saturated thickness over time are also significantly more likely to enroll. An additional 10% decrease in saturated thickness between 1980 and 2005 is predicted to increase the probability of enrollment by 0.89%. This suggests that not only is the level of groundwater availability important to landowners in deciding whether to retire a given well, but so too is the trajectory of water availability over time.

There are some differences in the coefficient estimates of the aquifer-specific variables in the models that utilize wells that are within eight miles of the North or South Fork of the Republican River. Although the change in saturated thickness no longer has a significant impact on enrollment, the coefficients on the conductivity and depth to water variables are significant and positive. The depth to groundwater variable is intuitive from the standpoint that wells with a greater depth to water are more expensive to operate. The fact that hydraulic conductivity is positively related to enrollment suggests that the degree to which the groundwater resource is shared by multiple users plays a role in the decision to enroll. The magnitude of the impact on enrollment of these two variables is similar to that of the saturated thickness variables. A 10% increase in conductivity or depth to water is predicted to increase enrollment by 0.90 and 0.25%, respectively.

The variables that are included to account for the variation in soil characteristics are also found to have an impact on enrollment. In the model that includes data from all eligible wells, only the dummy variable representing wells with a land capability class of I or II is found to have a significant impact on enrollment. Consistent with expectations, the marginal effect suggests that class I or II land has an enrollment rate that is 1.51% lower than land with a higher class. This outcome captures the fact that irrigated land that is more conducive to agricultural production is less likely to be retired. Restricting the sample to only wells within eight miles of the North or South Fork yields two additional variables that are found to have a significant impact on enrollment. In particular, irrigated corn yield and the available water capacity of the soil are both found to have a positive impact on enrollment. The positive coefficient estimate for the corn yield is somewhat counterintuitive, as it suggests that irrigated land with higher potential yield, controlling for land capability class, may be more likely to be retired. Since available water capacity is an important characteristic for dryland production, it appears that after controlling for the land capability class, landowners are more likely to enroll land that could be suitable for dryland production in the future.

6. Discussion and Conclusion

Although a number of previous economic studies have assessed the factors that influence participation in agricultural land retirement programs, this is the first effort, to our knowledge, to evaluate participation in a land retirement program that specifically targets water conservation. In addition, this is the only agricultural land retirement study that we are aware of that makes use of spatially explicit information on all land eligible for the program. By using well-specific information on the financial incentives offered, as well as the aquifer and soil characteristics of eligible wells in Colorado’s Republican River Basin, we estimate participation models that yield two important findings for conservation planning. The first finding relates to the relationship between the financial incentives that are offered and enrollment rates in the program. Specifically, we find that landowners enroll at higher rates as the financial incentives offered increase. The magnitude of the relationship, based on the marginal effect estimate from our full model, predicts that enrollment rates increase by 0.087 percentage points with a $10 increase in the net present value of the incentives offered.

Estimates of the marginal effects of incentives could be used by program administrators to develop predictions of how changes in the financial incentives that are offered will influence enrollment rates. In particular, our results imply that an increase of approximately $115 in the net present value of the incentives offered per acre is required to encourage an additional 1% of eligible wells to enroll in the program (nearly 3.5% of wells are currently enrolled in the program). As mentioned above, the USDA recently announced that annual payments offered to landowners that enroll in the program will increase by $65 per acre. This will serve to increase the net present value of the financial incentives for enrollment by more than $700 per acre. Based on our model results, we predict this will serve to increase enrollment in the program by more than 6%. Although the model results provide a general prediction, care must be taken in extrapolating the estimated marginal effects to discrete changes in incentives, which rely on an assumption of a constant, linear relationship between changes in incentives and enrollment.

Another important finding for policy makers interested in retiring irrigated land to conserve water is that several of the aquifer and soil variables affect the probability of enrollment. Although there are some differences in the statistical significance of the variables across models, we consistently find that wells with higher saturated thickness have a lower probability of enrollment in the program. In other words, landowners are more likely to retire a well if groundwater availability at that well location is limited. For example, based on the estimated marginal effect from the participation model, a 34.7 foot decrease in saturated thickness (approximately 25% of the mean saturated thickness) is predicted to increase enrollment by 1%. Decreasing saturated thickness by 34.7 feet is therefore predicted to have a similar effect on enrollment as increasing the net present value of the incentive payments by $115.

The implication of the fact that wells with lower saturated thickness are more likely to enroll in the program is an important consideration for future research. One could argue that primarily enrolling wells with limited saturated thickness does not provide a sufficient level of additionality toward water conservation goals, since these wells would have extracted lower volumes of water over time than wells that overlie areas with greater saturated thickness (Foster, Brozović, and Butler 2014). The wells that have been observed to enroll in the program had a mean saturated thickness of only 77.6 feet compared to 133.8 feet of saturated thickness for an average well in the study area. Using well pumping records for the Republican River Basin of Colorado for the years 2011–2014, we find that wells with a saturated thickness of between 72.6 and 82.6 feet pumped an average of 212.9 acre-feet of water per year (N = 261, std. dev. = 84.1), whereas wells with average saturated thickness (between 128.8 and 138.8 feet) pumped an average of 250.6 acre-feet of water per year (N = 98, std. dev. = 98.0). The difference in water use of 37.7 acre-feet per year is statistically different from zero (t-value = 3.37, p < 0.001). This simple calculation suggests that the groundwater conservation associated with the actual wells that enrolled was approximately 18% lower on an annual basis than if wells with saturated thickness equivalent to the basinwide average had enrolled in the program.

Retiring wells in areas with relatively low saturated thickness may, however, enhance efficiency if it means that the remaining groundwater is more likely to be used by wells with higher saturated thickness that also have a higher marginal product of water applied. Again, future research should consider how integrating landowner responsiveness to differences in saturated thickness could be used to enhance the efficiency of land retirement programs that seek to conserve water.

In addition to the aquifer characteristics, enrollment rates are also found to be responsive to soil characteristics of the land that is irrigated. In particular, our most robust finding is that land with soils that fall in land capability classes I or II are less likely to enroll in the program than land with higher classes. This suggests that enrollment of irrigated land is responsive both to the productivity characteristics of the land and to the productivity characteristics of the aquifer that it overlies. This may assuage the concern that land retirement does more harm than good for rural communities if it takes the most productive land out of production. Instead, since landowners with high-quality land are less likely to enroll, land retirement will tend to be more heavily distributed in portions of the program area with the lowest opportunity costs of enrollment.

In cases where policy makers have an objective of maximizing the number of irrigated acres or number of wells that are retired, the fact that enrollment rates respond to aquifer and soil characteristics implies that the incentives offered could be differentiated based on those characteristics. In particular, enrollment rates could be increased if wells that have lower saturated thickness and irrigate poorer soils could be offered lower incentives relative to wells that overlie a higher volume of water and more productive soil. In this way, the overall enrollment could be increased while holding the average incentive that is offered constant. Whether this increase in enrollment improves the efficiency of the program ultimately depends on how it impacts the long-run productivity of groundwater use and other benefits that come from reducing conjunctive use issues.

There are several aspects of this study that may limit our findings. We do not observe groundwater pumping prior to or during the enrollment period and are thus unable to determine which wells extracted the required volume of groundwater to be eligible for the program. In other words, some of the wells that we observed as not participating in the program may indeed not have been eligible for the program if they were not utilized sufficiently for irrigation in the years prior to the enrollment period. We also are not able to associate a particular groundwater well with the parcel of land that it irrigates. This means that we must rely on the soil characteristics at the well location rather than the characteristics of the land that is irrigated. Although it is very common for a well to be situated within the area that is irrigated, we are not able to observe cases where the soil characteristics at the well location differ substantially from the distribution of characteristics on the irrigated land. A final limitation of our study is that enrollment is modeled using static parameters for aquifer and soil characteristics in addition to prices and costs. In reality, the costs and benefits associated with retiring irrigated land change over time with changes in expectations related to future input and output prices, as well as changes in the productivity of the groundwater resource and climate. Given that the wells that enroll in the program are retired permanently, using a dynamic model that incorporates option value and captures the effect of changes in price expectations and net returns to irrigated production could yield additional understanding about participation behavior. Since we do not have sufficient data to capture these expectations over a sufficiently long interval, however, we have focused instead on static parameters.

There are several areas for additional research that could help to improve the social efficiency of programs that retire irrigated land with the objective of conserving groundwater resources. As described above, a better understanding of the relative social benefits associated with retiring wells with specific aquifer and soil characteristics could improve the targeting of outreach efforts and the potential for differentiated incentive rates that depend on a wider range of factors than distance from a surface water body.

Spatial patterns of well ownership could also be incorporated into the empirical participation model. Wells that are located in close proximity to retired wells will experience external benefits due to potentially slower rates of decline in saturated thickness and therefore relatively lower pumping costs. If well owners are able to internalize this spillover from retirement, enrollment rates may depend in predictable ways on spatial patterns of well ownership. Understanding how the external benefits of retiring irrigated acreage impacts the behavior of nearby groundwater users would also allow for an understanding of potential slippage from programs that retire irrigated land, building on work by Wu (2000) investigating slippage in the CRP. If groundwater users near retired wells respond to the higher levels of saturated thickness by extracting more groundwater than they otherwise would have, then some of the groundwater conservation benefits associated with retiring irrigated land are dissipated.

Finally, producers in many regions have the option to enroll irrigated land in other programs such as the Environmental Quality Incentives Program (EQIP) and the Agricultural Water Enhancement Program (AWEP), in addition to the CREP. These programs restrict producers from irrigation but allow dryland production during the contract period, although the financial incentives that are offered also tend to be lower. Rather than estimating bivariate models of enrollment, models that account for a wider range of land use and program participation options (Claassen et al. 2017) could yield additional insight into the factors that drive enrollment behavior.