In recent years there has been increasing regulation of agricultural water use in order to reduce transboundary and environmental water conflicts. Effective policy analysis needs to have tools to estimate correctly the value of irrigation water. Irrigating land increases crop yields and this higher profitability should be capitalized into the sales price of the land. For irrigation that depends on surface water rights, studies have found this to be the case (Xu et al. 1993, Faux and Perry 1999). However, studies that have analyzed the value of groundwater in irrigation have found mixed results. Hartman and Taylor (1989) and Sunderland, Libbin and Torell (1987) find that groundwater irrigation has no significant effect on land prices; Torrell et al. (1990) find a significant positive effect of groundwater in irrigation. One explanation is that in areas where groundwater use is not restricted there is the option to implement irrigation in the future and thus the presence of groundwater irrigation may not have a large effect on the sales price. Consistent with this idea of option value, Petrie and Taylor (2007) look at differences in land values before and after a moratorium on water-use permits and find that permits add value to agricultural land only after the restriction is in place. An additional econometric issue is that the decision to irrigate is not random but is based on the underlying characteristics of the land. Thus hedonic estimates of the value of irrigation rights may be biased. In this paper we analyze the value of groundwater in an area with pumping restrictions using both a standard hedonic model and a propensity score matching model. We use a geospatial database from Chase County, Nebraska that includes arms length sales, tax assessor’s data, hydrologic and climatic variables. We find that per acre values of groundwater irrigation are 15 percent higher using the propensity score method compared to the hedonic model. This result is driven in large part by the preferential adoption of irrigation on intermediate quality land. An important implication for policy is that hedonic estimates of the value of groundwater in irrigation may underestimate the cost, to both farmers and the government, of future water use reductions. Our study area is part of the Republican River Basin in Nebraska. The Basin has been the source of litigation between Colorado, Kansas and Nebraska. In 2002, the Supreme Court decided that groundwater pumping by Nebraska caused reduced instream flows in Kansas. As a result, groundwater management districts in Nebraska were forced to introduce a variety of restrictions: moratoria on new wells (introduced in 1999), metering of existing ones, and volumetric pumping restrictions. The data for this research include all agricultural parcels in Chase County, Nebraska that sold between 2000 and 2008, obtained from the Chase County tax assessor’s website. Each parcel contains sale prices along with the new ownership and sales date. In addition, the presence of outbuildings and the square footage and age of residences are included. Each observation also includes the distribution of agricultural land (irrigated, dryland or grassland) and four soil quality types in each type of land, resulting in a total of twelve soil classes. We also assembled georeferenced data on estimated depth to water, the rate at which a well can pump water, precipitation and growing degree days. After excluding the non arms-length sales transactions, outliers where the sale price per acre was greater than $6,000, and parcels of less than four acres, 330 observations were left. In the first estimation, we use physical, geophysical, sales, and structure data to characterize the sales price. We estimate an Ordinary Least Squares Regression model in which the dependent variable is the sales price per acre regressed on the other variables. Consistent with Petrie and Taylor (2007), our results show that irrigated land sells for a higher price than equivalent dryland or grassland (ceteris paribus). For the period between 2000 and 2008, during which the well drilling moratorium was in place, we find that the price of irrigated land was about $712 per acre greater than that of non irrigated land. Second, we estimate a propensity score model which involves two steps. First, we estimate the probability of a parcel of land being irrigated using a probit model. An interesting result from the probit regression is that farmers are more likely to irrigate on lands that are of intermediate quality (Soil 2 and Soil 3), suggesting that irrigation is a land-quality augmenting technology. This result is similar to Lichtenberg’s (1989) analysis in the region, but unlike that paper our results suggest that irrigation is less likely on the lowest quality land. In the second step we match pairs of non irrigated and irrigated parcels that have the same probability of being irrigated and analyze whether there is a difference in the sales price. Results show that once again there is a positive and significant value associated with groundwater irrigation rights, but the difference is larger than the hedonic analysis at $839. This suggests that if selection into irrigation is not accounted for, the price of the land is undervalued by over 15 percent. These results have important policy implications in areas such as Nebraska where there is debate on how to reduce water use effectively. The results of our paper suggest that future reduction may be costlier than previous research suggests and that the standard hedonic method may not be best in estimating the value of water. Also, the reductions will be more expensive to both the government and farmer. In Nebraska, the government has tried to reduce water use by buying permits from farmers. There has been a lot of resistance to this and one explanation for this may be that the government is not offering enough compensation for the permits.