Competition for Water: Farming vs. Fracking Claudia Hitaj, Jeremy G. Weber, and Andrew Boslett United States Association for Energy Economics North American Conference October 25-28, 2015 Pittsburgh, PA Fracking, which became commercially viable in the early 2000s, uses several million gallons of water per well. Oil and gas production from shale wells declines rapidly, by about 70% by the end of the first year. This means new wells must be drilled continually in order to keep up the supply of oil or gas from shale formations. Between 2008 and 2013 about 40,000 new wells were drilled each year. Jeremy Weber, Andrew Boslett, and I are studying the effect of the recent growth in oil and gas extraction on agriculture, which is the most water-intensive industry in most regions. I became interested in this topic, after I read conflicting statements from the energy and agricultural industry on the severity of this issue. *The views are those of the authors and should not be attributed to the USDA or the Economic Research Service

Fracking and Farming in the News “When drought occurs, fracking and farming collide” (Denver Post, Feb 2014) “Kale or fracking? Farmers and corporations fight it out for water” (The Guardian, Nov 2014) Hydraulic Fracturing & Water Stress, Ceres (2014) Energy companies and Environmental groups are emphasizing different facts. It is immediately clear that if there is any effect from fracking on water use in agriculture, the effects will be localized. Consider Kern County, California. It is home to the North Belridge and South Belridge oil fields, the country’s sixth largest oil areas. A lot of the production in the Belridge fields has come from fracking, or hydraulic fracturing, a process that involves injecting a mixture of water, sand and chemicals into shale rocks containing oil and gas to break them open, enabling the well operators to extract them.  The region, near Bakersfield, also is home to half of the country’s carrot crop and 40% of its pistachio production, Lee says. These two industries are in direct competition for water. Last summer, Kern County was in crisis mode, getting only about a third of the water it had expected from the California State Water Project. So, who gets that water: the Shell/ExxonMobil joint venture now running Belridge, or the carrot and pistachio farmers?  EnergyInDepth.org (2014)

Motivation Academic research on shale development and water has focused on water quality (Olmstead et al., 2013; Vidic et al., 2013) We focus on water quantity: does fracking crowd out water use in agriculture? Fracking uses several million gallons per well Fracking in Texas’ Eagle Ford Shale can at times account for nearly 90% of local water use (Nicot and Scanlon, 2012) Local water stress In 2013, on average about 20-30 wells per county in drilling regions of Arkansas, Kansas, Oklahoma, 50 in Texas, 80 in Colorado, and 110 in North Dakota. Water, chemicals, and sand are pumped into the well to unlock the hydrocarbons trapped in shale formations by opening cracks (fractures) in the rock and allowing natural gas to flow from the shale into the well. • When used in conjunction with horizontal drilling, hydraulic fracturing enables gas producers to extract shale gas economically. • Without these techniques, natural gas does not flow to the well rapidly, and commercial quantities cannot be produced from shale 2 – 10 million gallons of water per well These next maps look at the geography of shale production, water use, and irrigation across regions.

Shale gas: - Barnett Fayetteville Haynesville Marcellus Eagle Ford Tight oil: Bakken Permian Eagle Ford in TX and Bakken in Montana and North Dakota, and the Permian in TX are the most important tight oil plays. Marcellus, Haynesville, Eagle Ford, Fayetteville, and Barnett are the most important shale gas plays. Type of play, depth, gas only or oil and gas, all play a role in how much water is needed for fracking a well.

Oil and Gas Extraction: Water Use Varies Within Plays This includes vertical wells, which require less water. Bakken: ranges from 1.3 to 7.7 gal/MMBtu oil (south-eastern part to the northwestern part) or 0.82 to 1.89 million gallons per well Eagle Ford: 10.1 gal/MMBtu of oil versus 4.1 gal/MMBtu of gas Water use depends on a variety of factors, such as depth, type of rock formation, whether there is oil, dry gas, or a combination of both High water use per well in Texas, Oklahoma, Kansas Arkansas, Louisiana Wyoming, Colorado, ND and Montana Map created using data from USGS

Acres of Irrigated Harvested Cropland as Percent of All Harvested Cropland Acreage: 2012 Most irrigation in western US Some overlap: Arkansas & Louisiana Texas, Oklahoma, and Kansas Colorado, Wyoming U.S. Department of Agriculture, National Agricultural Statistics Service

Theory Growth in water demand not necessarily a problem if institutions account for water scarcity But they may not (e.g. rule of capture), and fracking may cause further overuse A lack of crowding out may reflect Fracking uses too little water to affect prices Poorly functioning water markets (e.g. prices and use don’t change) A lack of crowding out could be good or bad. For one, it could mean that too little water is used in fracking to affect water prices. On the other hand, it could mean that water policy does not account for the scarcity of water (e.g. rule of capture), in which case people just pump more in the present (hence no change in the price of water), to the detriment of the future. Texas groundwater belongs to the landowner. Groundwater is governed by the rule of capture, which grants landowners the right to capture the water beneath their property. The landowners do not own the water but have a right only to pump and capture whatever water is available, regardless of the effects of that pumping on neighboring wells. Surface water, on the other hand, belongs to the state of Texas. It can be used by a landowner only with the state's permission. We could be pumping too much groundwater to the detriment of the future.

Data Farm-level Farm and Ranch Irrigation Survey (FRIS) from 2003, 2008, and 2013 County-level Water use per well across shale plays (USGS) Wells drilled (DrillingInfo) Palmer modified drought index (NOAA) FRIS is farm-level. Quantity of water used, acres irrigated, investment in irrigation equipment HUC – hydrologic unit code, 8 digit, converted to county-level The Palmer Drought Severity Index (PDSI) uses readily available temperature and precipitation data to estimate relative dryness. It is a standardized index that spans -10 (dry) to +10 (wet). - See more at: https://climatedataguide.ucar.edu/climate-data/palmer-drought-severity-index-pdsi#sthash.50WsRe5r.dpuf

Methodology Dependent variables Quantity of water for irrigation on farm (gallons and gallons/acre), and price of water Independent variables Quantity of water used for fracking normalized by county area Drought index and drought index squared Models Fixed effects at the farm level First difference with state and state*year FE Fixed effect and first difference controls for time-invariant unobserved characteristics at the farm level Price of water is determined as the cost of off-farm surface water divided by the quantity of off-farm surface water Control for time-invariant farm-level characteristics with fixed effects or by differecing. Almost 50% of farms were surveyed at least twice, 13% were surveyed in all three years

Fixed Effects: Irrigation Water (Gallons) Dependent variable: Water for irrigation (gallons) All shale states Kansas Louisiana Montana Water for fracking -0.060 -2.151* 0.457** 3.937** (gallons/acres in county) (0.057) (1.231) (0.229) (1.717) Drought index 134.033*** 63.942 6.550 67.742 (24.461) (113.787) (172.173) (54.095) Drought index squared -8.334*** -4.517 0.710 -4.884 (1.514) (7.537) (11.994) (2.990) Constant -395.947*** -19.364 -12.579 -163.357   (92.990) (409.499) (603.462) (220.331) Number of observations 28,232 1,594 2,350 1,791 Adjusted R2 0.006 0.008 0.007 0.003 Not unsurprisingly there is no uniform effect across all regions, since water supply varies. However, we were surprised to see a significant positive effect as well. We think this could be a wealth effect, farmers receive royalties and invest them in irrigation equipment.

Fixed Effects: Irrigation Water (Gallons/Acre) Dependent variable: Water for irrigation (gallons/acre) All shale states Arkansas Oklahoma Texas Pennsylvania Water for fracking 20.685 -35.795** -55.237** 42.596* 6.562** (gallons/acre in county) (18.419) (17.880) (23.980) (23.772) (3.152) Drought index 59,878*** -214,396 -8,756 117,784*** 75,212* (12,327) (144,320) (16,328) (45,196) (39,236) Drought index squared -4,042*** 6,618 525 -8,933** -4,220** (739) (8,588) (951) (3,751) (2,117) Constant -1,208 1,613,224*** 104,249 -243,039* -308,328*   (47,987) (604,335) (68,281) (133,306) (180,241) No. of observations 21,399 1,709 1,098 2,154 1,179 Adjusted R2 0.014 0.152 0.005 0.057 0.019

First Difference Results D.Water for irrigation (gallons) D.Water for irrigation (gal/acre) Stacked 2008-2013 2003-2013 D.Water used for fracking -0.043** -0.023 -0.044 6.508 -1.012 -17.626 (gal/acres in county) (0.022) (0.026) (0.030) (8.724) (12.227) (13.600) D.Drought index -19.6 -44.5** -111*** 32,816*** 28,347*** 16,100* (16.044) (18.385) (21.302) (6,546) (8,347) (9,222) D.Drought index squared 0.317 1.943 5.808*** -2,413*** -2,123*** -1,180* (1.051) (1.209) (1.406) (430.951) (547.693) (610.168) Indicator if in 03-08 148*** 15,842** (17.0) (6,236) No. of observations 28,237 12,221 16,471 6,377 6,282 Adjusted R-squared 0.016 0.007 0.005 0.014 0.009 D. stands for the 5- or 10-year difference

Fixed Effects: Water Price Dependent variable: All shale states Wyoming Texas Water price ($/million gallons) Water for fracking -0.00036 0.00242 0.00308 (gallons/acres in county) (0.001) (0.009) (0.002) Drought index -0.428 -0.077 -1.608*** (0.272) (1.421) (0.293) Drought index squared 0.026 0.009 0.107*** (0.018) (0.118) (0.026) Constant 5.620*** 3.287 10.324***   (0.978) (3.951) (0.845) Number of observations 3,525 423 214 Adjusted R-squared 0.022 -0.001 0.319

Preliminary Results: Mixed In this specification, can detect: No effect of fracking on the cost of surface water to farmers No nation-wide effect of fracking on water use for irrigation Effect varies across regions Negative effect Arkansas (Fayetteville), Oklahoma (Woodford), Kansas (Woodford, Excello) Positive effect Louisiana (Haynesville), Montana (Bakken), Texas (Eagle Ford, Barnett, Haynesville), Pennsylvania (Marcellus) What factors are we missing? Negative effect Woodford: tight oil + shale gas Fayetteville: shale gas Kansas: ? Woodford (tight oil + shale gas) and Excello-Mulky (shale gas?) Positive Effect Eagle Ford: tight oil Haynesville: shale gas Barnett: shale gas Marcellus: shale gas Bakken: tight oil

Limitations Do not account for: Production changes within a farm over time Switching to a different crop Improving efficiency of irrigation machinery Different water markets and sources of water Reuse of fracking water Larger farms are more likely to be part of the survey multiple times Water supply issues

Summary Fracking can have a negative or positive effect on irrigation in agriculture Competition for water Wealth effect for farmers through increased economic activity or royalties A lack of crowding out may reflect Poorly functioning water markets Isolated water markets Fracking uses too little water to affect prices Future work: Test additional model specifications and samples Identify the role of different water markets difficult to distinguish between an agriculture-wide effect and an irrigation-specific effect With new data, we can test for the impact of royalties on irrigation

Shale Production

U.S. Drought Monitor, Week of April 14, 2015 It’s important to control for drought, since it could exacerbate any competition for water. Drought conditions change, but here Texas, Oklahoma, Kansas are affected droughtmonitor.unl.edu

Water, chemicals, and sand are pumped into the well to unlock the hydrocarbons trapped in shale formations by opening cracks (fractures) in the rock and allowing natural gas to flow from the shale into the well. • When used in conjunction with horizontal drilling, hydraulic fracturing enables gas producers to extract shale gas economically. • Without these techniques, natural gas does not flow to the well rapidly, and commercial quantities cannot be produced from shale 2 – 10 million gallons of water per well ProPublica

Fixed Effects: Shale Coverage Ag-wide effect Irrigated acres Value of ag products Non-irrigated harv. acres Percent of county covering a tight or shale formation 0.141*** 1,427.188 0.184*** (0.027) (1,116.857) (0.055) (year=1997)*percent tight/shale -0.068*** -80.693 -0.117*** (0.006) (115.671) (0.012) (year=2002)*percent tight/shale -0.176*** -44.888 -0.202*** (0.008) (124.960) (0.017) (year=2007)*percent tight/shale -0.263*** -733.568*** -0.274*** (0.009) (229.356) (0.020) (year=2012)*percent tight/shale -0.300*** -1,248.265*** -0.535*** (0.011) (284.409) (0.025) For some reason we see a negative effect of being in a shale area on irrigated acres, even before shale development took place! The large number of small farms could be messing things up Observations: more than 2 million Non-irrigated harvested acreage does not include pastureland or non-harvested land

Fixed Effects: Energy Production Ag-wide effect Irrigated acres Value of ag products Non-irrigated harv. acres Btu of energy (oil + gas) per county area in acres 0.027*** -421.369** 0.035*** (0.008) (202.283) (year=2007)*Btu energy -0.021*** -12.572 -0.051*** (0.007) (92.212) (year=2012)*Btu energy -0.029*** -114.694 -0.064*** (164.631) Observations: 1.1 million Non-irrigated harvested acreage does not include pastureland or non-harvested land

First Differencing Dep. Var.: D.Irrigated Acres Shale states Irrigation areas D.Btu Energy (02-07 and 07-12) -0.001 -0.012 (0.002) (0.015) D.Btu Energy (07-12) -0.003 -0.133*** (0.003) (0.032) D.Btu Energy (02-12) -0.006* -0.091*** (0.027) !!! Mention that this is a table showing the regression coefficients from SIX different regressions. We did the stacked difference and the two sets of 5 year differences separately.!!! Observations: Stacked difference: 860,000 for shale states and 174,000 for irrigation areas Difference 07-12: 280,000 for shale states and 58,000 for irrigation areas Difference 02-12: 250,000 for shale states and 52,000 for irrigation areas With First Differencing we see an effect only in the sample of farms in irrigation areas Irrigation areas share of farmland irrigated in 1997 by county greater than 5% Shale states 5 = Arkansas 8 = Colorado 20 = Kansas 22 = Louisiana 30 = Montana 31 = Nebraska 32 = Nevada 35 = New Mexico 36 = New York 38 = North Dakota 39 = Ohio 40 = Oklahoma 42 = Pennsylvania 46 = South Dakota 48 = Texas 49 = Utah 54 = West Virginia 56 = Wyoming

Oil/Gas Production Over Time (County-Level) Dep. Var.: Oil & Gas Production (Btu/acre) Shale states WY, CO, TX Percent of county covering either a tight or shale formation 177.711*** 287.924*** (47.167) (110.377) (year=2007)*percent tight/shale 171.040*** 308.480*** (45.211) (114.166) (year=2012)*percent tight/shale 555.157*** 757.189*** (118.897) (229.693) Observations 3,184 860

Instrumental Variables (County-Level) Dependent variable D.Log Irrigated Acres D.Share Irrigated Acres D.Log Non-Irrigated Cropland D.Log Value Ag Production D.Log Value/Acre D.Oil/Gas production 0.520 -1.640 -0.030 -0.805*** -0.036 (Btu/acre) (0.553) (2.075) (0.294) (0.311) (0.194) D.Drought index 0.018* 0.090** 0.048*** 0.019*** 0.007 (0.010) (0.038) (0.009) (0.005) D.Drought index squared -0.001 0.004 -0.001* -0.001*** 0.000 (0.001) (0.002) (0.000) Constant 0.068 2.686*** 0.047 0.286*** -0.007   (0.109) (0.595) (0.049) (0.045) (0.032) Number of observations 2,055 2,054 1,842 Adjusted R2 0.019 0.123 0.127 0.076 0.463 Instrument for oil/gas production per acre with Percent of county covering a tight or shale formation Instrumenting for oil & gas production per acre with percent of county covering a tight/shale formation Non-irrigated harvested acreage does not include pastureland or non-harvested land

Conclusions Mixed results at the national level across the different specifications: Some evidence of an impact of fracking on water use, particularly in irrigation areas (localized impacts) Negative agriculture-wide effect A lack of crowding out may reflect Poorly functioning water markets Fracking uses too little water to affect prices Future work: Focus on the intensive margin using FRIS data Test additional model specifications and samples Identify the role of different water markets Interact rainfall and the effect of fracking difficult to distinguish between an agriculture-wide effect and an irrigation-specific effect