Shale Gas Development and Housing Values Over a Decade: Evidence from the Barnett Shale

Transcription

1 Shale Gas Development and Housing Values Over a Decade: Evidence from the Barnett Shale Jeremy G. Weber (USDA/Economic Research Service) Wesley Burnett (West Virginia University) Irene M. Xiarchos (USDA/Office of Energy Policy) International Association for Energy Economics, Energy and the Environment Conference, June 15-18, NY, NY *The views are those of the authors and should not be attributed to the USDA or the Economic Research Service

2 Shale Development and Home Values Quality of life concerns have led to moratoriums on hydraulic fracturing. New York, North Carolina, Maryland, and Vermont Proposed moratoriums elsewhere: California, Colorado, and Illinois Concerns Macro: expanded fossil fuel extraction slows the shift towards renewables Micro: groundwater contamination, air pollution, truck traffic and road deterioration, waste disposal, and surface spills. Will drilling leave a mountain-top-mining-like legacy in my area?

3 The Empirical Question What happens to the value of the typical house in an area experiencing development? Strengths : summarizes the net local effect. Weakness : mask heterogeneous effects How does the effect change over time? Impacts in 4 phases of development

4 Region: Dallas Fort Worth (DFW) & Barnett Shale Years: 1997 to : pre-drilling : slow rise in drilling : boom in drilling : modest drilling Outcome: Monthly Changes of Zillow Housing Value Index (ZHVI) for DFW zip codes The ZHVI tracks S&P/Case-Shiller index well Model Δ ln ZHVI it = β 0 +β 1 Shale i + β p Shale i Period p(t) + f Month m(t), Period p(t), t t + ε it

5 Drilling Permits Submitted Why the Barnett? More wells, smaller area, longer history than other plays Key shale counties: Denton, Johnson, Parker, and Tarrant Source: Texas Railroad Commission.

6 Source: Energy Information Administration.

7

8 Shale/nonshale status Identification based purely on geological endowments Sample processing improves comparability and robustness to misspecification (Crump et al, 2008) Processing using the propensity score matching, without replacement Keeps all shale zip codes and drops most dissimliar nonshale zip codes Trimming using optimal cutoffs (approx to 0.95) drops the most extreme shale and nonshale zip codes

9 Shale and Nonshale Zips Average normalized difference: Full sample: 0.169, Matched: 0.096, Trimmed: Normalized Differences Variable Full Matched Trimmed Race, Age, Education Share white Share age Share age Share age 60 or older Share with some college or more Share 20K-40K Income Share 40K-75K Share 75K or more Median household income Share urban Urbanization Population density Real Estate Share vacant Median year built Median real estate taxes

10 What do we expect? The zip code ZHVI (average value) is affected by (-) by disamenities and health/environmental risks (+) by economic growth and improved local public finances NET EFFECT (+, or -) changes with time: Which term dominates is an empirical question. The net effect may be positive initially but turn negative as wells age or vice versa

11 Permits Submitted Difference in Mean Housing Valus, Shale and Nonshale Zip Codes ($/house) Drilling Permits and Housing Values, ,000 25,000 4,500 4,000 3,500 3,000 2,500 Submitted Well Permits: Shale counties Housing Values: Difference between Shale and Nonshale Zip Codes 20,000 15,000 2,000 10,000 1,500 1,000 5,

12 Mean Difference in Monthly Appreciation (%): Shale - Nonshale Differences in Appreciation by Period Full Sample Matched Trimmed *Mean differences are conditional on time trends and control variables

13 Total Appreciation, Y=ln(ZHVI_2013)-ln(ZHVI_1997) Full Sample Matched Trimmed ZHVI, Shale 0.052** 0.057** 0.058** (0.022) (0.024) (0.023) Intercept (7.167) (9.253) (8.984) Control for initial characteristics Y Y Y Number of observations Adjusted R *Estimates of total appreciation based on Census data give estimates similar to ZHVI

14 Drilling Intensity Number of wells drilled in a zip code per square km Using 29 shale zip codes: one more well per square km correlated with 1.6 percentage point less in appreciation from Reflects disamenities or endogeneity of well placement?

15 Why Greater Appreciation in Shale Areas? Improved local finances Subsurface resources taxed as property in Texas Property taxes fund local governments and schools improving home values Model: y it = α 0 + α 1 Shale i + α p Shale i Period + η it y it : The tax base ($) per enrolled student Land, Commercial, Oil and Gas, Residential, Total School district finances per enrolled student Local, State, Total revenues, Operating expenditures, and fund balances

16

17 Changes in Tax Base per Enrolled Student Total Tax Base Commercial Land Oil and Gas Residential Other Shale 21,594 24,519* 1,744 5,236*** -9, (27,930) (14,444) (6,087) (1,478) (17,670) (502) Shale x Period ,363 2,951 1,713 11,743* -12,563* 519 (11,835) (4,099) (2,339) (6,382) (7,238) (329) Shale x Period ,729*** 28,407** 8,473 53,319*** -10, (25,028) (11,816) (6,072) (15,912) (11,089) (424) Shale x Period ,899*** 55,173*** 15,906** 82,008*** 18,129* 683 (30,474) (14,451) (7,937) (17,310) (10,735) (444) Period ,892*** 10,456*** 7,959*** -8 43,214*** 1,272*** (7,228) (2,196) (1,137) (7) (6,572) (200) Period ,176*** 25,489*** 18,378*** 39 83,201*** 1,068*** (11,380) (4,562) (3,331) (32) (9,335) (340) Period ,852*** 26,915*** 16,003*** 63 77,872*** -2 (9,488) (4,441) (4,434) (48) (7,124) (345) Intercept 189,740*** 43,061*** 40,395*** 24** 103,353*** 2,906*** (20,323) (7,365) (3,263) (11) (14,704) (357) Number of school districts Number of observations 1,176 1,176 1,176 1,176 1,176 1,176 Adjusted R-squared Source: data from the Texas Education Agency

18 Changes in School Finances per Student Local Tax Revenue State Revenue Total Revenue Operating Expenses Source: data from the Texas Education Agency Fund Balance Shale (346) (317) (178) (209) (184) Shale x Period (147) (135) (126) (171) (143) Shale x Period ** -409** 310* 324* 523*** (245) (177) (184) (180) (201) Shale x Period ,590*** -1,190*** 595*** 460** 454* (316) (242) (207) (198) (237) Period *** 461*** 1,853*** (92) (101) (96) (160) (108) Period ,476*** 598*** 2,940*** (126) (120) (112) (167) (170) Period ,516*** 1,611*** 3,948*** 1,038*** 322** (131) (152) (115) (165) (164) Intercept 2,361*** 3,716*** 6,501*** 6,431*** 1,161*** (238) (223) (124) (175) (151) School districts Observations 1,176 1,176 1,176 1,176 1,176 Adjusted R-squared

19 School District Findings: Implications Improved school finances account for much of the greater appreciation of shale zip codes (75-90%) By 2011 total school revenues p.s. had increased about $600 more in shale zip codes (9 % from 1997). Assuming all revenues spent and housing elasticity of 0.5, housing appreciation is 4.5 % compared to total appreciation of 5-6 %.

20 Caveats Housing values may not reflect the cost of all disamenities. Currie et al.(2012): health effects of toxic plant extend over a larger area Risk perceptions and housing capitalization reflect available information. Sanders (2012): households underestimate risk prior to information provision Findings from other regions/plays may be different Different priors and risks TX: gas rights taxable property; PA: not included Suburban versus rural

21 Summary Compared to nonshale zip codes, shale zip codes appreciated The same prior to development More during peak development Less after peak development The net effect ( ) is positive: 5-6 % points increase. In part reflects improved local public finances: Value oil and gas rights expanded tax base by $82,000 per student For some (or most) residents it s possible drilling created disamenities Within shale zip codes an extra well per square kilometer associated with a 1.6 % point in appreciation over the study period.

22 Heteroskedastic robust standard errors in parenthesis. Appreciation by Period Full Sample Matched Trimmed Shale (0.0003) (0.0004) (0.0004) Shale x Period * ** (0.0004) (0.0006) (0.0005) Shale x Period *** *** *** (0.0005) (0.0005) (0.0005) Shale x Period * ** * (0.0004) (0.0005) (0.0005) Period (0.0085) (0.0107) (0.0102) Period * (0.0158) (0.0188) (0.0192) Period *** *** *** (0.0203) (0.0221) (0.0225) Intercept (0.0357) (0.0397) (0.0406) Control for initial charactersitics Y Y Y Month dummies Y Y Y Time trend x period dummies Y Y Y Number of zip codes Number of observations 15,200 10,800 10,400 Adjusted R-squared