A recent research paper entitled Natural Gas Operations and Infant Health, by the Department of Applied Economics and Management Ph.D. student Elaine L. Hill, attempted to identify a causal relationship between natural gas operations and infant health.  While admittedly a “work in progress,” the report largely ignored other factors that should have been addressed in any objective study.  Ms. Hill also relies upon other biased work (e.g., Bamberger and Oswald) to build the case for her own conclusions and pushes forward a great deal of unsubstantiated speculation by other like-minded researchers.  These include discredited sources such as Al Armendariz (author of what she refers to as the Ramon Alvarez report, but he was the only recipient) and Theo Colborn.  Finally, the report is filled with unsourced rhetorical statements that illustrate its unreliability.

Problems with the Paper’s Premise

The empirical method Hill employed was the differences-in-differences approach, which examines the differences in the predetermined birth measures over time (before and after gas well development) and the distance from a permitted developing gas well, in a population of mothers who had singleton births.  She utilized birth measures, including low birth weight (LBW), AGPAR scores and premature births, as health indicators.  The author also examined some confounding relationships that could have an impact on the results in an attempt to ensure the research data would be considered valid.

The hypothesis the paper set out to prove was that mothers who delivered singletons, and lived close to natural gas production, had poor standard birth measures (low birth weight, premature birth, small for gestational age data and 5 minute APGAR scores).  Other data was obtained from Pennsylvania Department of Environmental Protection (DEP) natural gas well records and vital statistics sources.  The distances from the wells examined were within radii of 1.5 km, 2 km and 2.5 km.  A population size of 1,069,699 over a seven year period (2003 to 2010) was used.  The sample size for mothers exposed to gas development within 2.5 km was 2,437.

The report’s introduction immediately signaled where it was going, with a discussion of natural gas industry exemption from the Safe Drinking Water Act (SDWA) and the Clean Water Act (CWA), and no concurrent mention of extensive state regulation.  It also included a number of baseless suppositions regarding natural gas development, indicating, for example, “serious environmental and health concerns have emerged regarding that may outweigh the perceived benefits of the technique.”  There is a source cited for this assertion (the Colorado Oil & Gas Conservation Commission, 2009) but when it is checked the word “health” doesn’t appear and there is nothing else even remotely close to Hill’s statement to be found there.  So much for sourcing.

Hill also states the purpose of the research was to “shed light on the matter” by investigating the causal relationship between natural gas development and infant health in Pennsylvania.  She also suggested her research could illuminate the adverse effects of pollution on fetal health, noting newborns make good research subjects because of the immediate cause and effect relationships between exposure and health effect (infant health indicators).  She also states her work could prove the EPA study Evaluation of Impacts to Underground Sources of Drinking Water by Hydraulic Fracturing (2004) was erroneous when she says it concluded “natural gas development” was safe.  She further claimed “growing anecdotal evidence is suggesting otherwise” and “the political environment from 2005 until 2010 allowed for nationwide regulatory exemptions.”

While it is true that pollution is a risk factor for LBW, most of the studies Hill referenced examined the LBW in relation to measurable pollutant levels.  The levels were then used to establish definitive and completed pathways and cause and effect relationships.  This study does not establish any pathways.  Rather, it only offers a vague causal relationships between air and water pollution and natural gas development that may not exist at all. The supposed relationships developed in Hill’s research rely upon very incomplete comparisons.  Consider the following examples.

Does Low Birth Weight Indicate Pollution or Something Else?

There are many causes of Low Birth Weight (LBW).  However, the author focused on natural gas development related activities and a short list of socioeconomic variables.  The statistical methods selected for this research, the differences-in-differences approach which measures the differences between the outcomes within the group (before and after gas development activities) over time is a commonly used method for these types of analysis.  However, its major drawback is endogeneity bias.  Such bias is found with this study as many confounders (variables) were not taken into account.  These include several other known causes of LBW (e.g., birth defects, maternal chronic health effects, smoking, alcohol and illicit drugs, infections, placental problems, inadequate maternal weight gain, inadequate prenatal care and other socioeconomic factors).  The lack of data on the types and the exposure levels to chemicals, in-migration in cases of residential mobility, medical history, occupational exposure to chemicals, dietary choices, lifestyle choices and the like are all factors not adequately examined.

Other studies that have attempted to establish a relationship between LBW and pollution have run into similar problems. This study conducted in Chicago found mixed results because of the uncertainty of exposure data associated with the distance from the monitoring source.  A similar style effect can be seen when one looks at the inconsistency of the results in the Cornell study.  One would expect the change in LBW to be directly related to the distance from natural gas production due to a pollution gradient, but the study results indicate great inconsistency; a 21% increase within the area closest to the well (1 km radius), then up to 26% at 1.5 km and then down to 25% at 2.5 km.  These numbers do not correlate with the way chemicals decapitate spatially in a pollution gradient and indicate a serious deficiency in the number of factors considered.  There are obviously other parameters influencing the outcome, factors Hill has not analyzed.

Source: Wall Street Journal

LBW numbers also exhibit an upward trend regardless of natural gas development patterns, further indicating other factors are involved.  This article discusses the trend.  The authors note there has been a steady increase each year since 1990 and gives reasons why, so we would expect some increases in the numbers of LBW in Pennsylvania regardless of natural gas.  Mean birth weight among all term singleton neonates decreased by 52 grams or an average of 4.25 grams per year during the 16-year study period, so this provides some indication of what might be expected.  Importantly, there are no definitive answers to this relatively new trend but we do know it is happening everywhere and started well before unconventional natural gas development began in earnest.  Ms. Hill appears to have attempted to find an answer that fit her ideology but her ideology doesn’t fit the facts.

Fluctuations within statewide LBW patterns are also to be expected around the average.  Additionally, these clusters could be random (entirely random or centered around a source or effect).  Think of the occurrence of LBW clusters in this way: if you drew random circles all around Pennsylvania, there would be some circles with no cases of LBW while there will be others where cases may cluster.  This random cluster phenomenon (also referred to as the Texas Sharpshooter Fallacy) has been seen when people look for causal relationships and completely overlook the possibility of randomness.

This study also overlooks the importance of exposure pathways, dose/response relationships and disease progression.  Health effects such as LBW do not happen instantaneously.  There is a continuum of effects that must occur for LBW to happen.  Therefore, exposure to a pollutant doesn’t always lead to LBW.  The exposure has to lead to an internal dose high enough to be biologically effective in causing an alteration in a body system or function.  Once an alteration happens, the body normally will first work to correct it.  If it is unable to correct the deficiency, then a health effect will be observed.  If there is no exposure above health threshold values, then there is no health effect.  Hill has overlooked all of this.

Generalizing Won’t Produce a Valid Answer to a Specific Question

Hill’s report is replete with generalizations served up as pseudo-scientific answers to specific questions.  She criticizes the 2004 EPA study as flawed, contending natural gas operations contribute to both air and water pollution which, according to her paper, are the fundamental causes for the LBW of infants.  Nonetheless, purported pollution evidence was reviewed by the University of Texas Energy Institute, which determined there is no water pollution caused by fracturing fluids in drinking water that could lead to widespread health problems. It found pollution problems that existed were related to faulty well casings that had caused methane migration or involved mechanical failure.  Both of these problems are limited and solvable.

Hill boastfully suggests her research paper could also be used to elucidate the need to reverse the so called “exemption” of the industry from the SDWA and the CWA.  This exemption has been explained and discussed here in great detail.  The short of it; the Energy Policy Act of 2005 served to codify the fact Congress never intended hydraulic fracturing to be regulated under the SDWA.  Hydraulic fracturing does NOT inject fracturing fluids into drinking water aquifers, but, rather, injects the fluids into the formations containing the targeted oil and gas, so there is no relationship.  The technique of hydraulic fracturing was in existence before the SDWA and was, therefore, never regulated under the SDWA.  There is correspondence from the EPA itself which verify this history.  The so called “Halliburton loophole” served only to restate what was and continues to be the legal acceptable practice.

The source of contention is the fact the SDWA does include Underground Injection Control (UIC) regulations that regulate underground wells. However, just because the language in the 2004 Energy Act exists, it does not mean the industry is not subject to UIC regulations. If you look at New York State’s SGEIS, it clearly discusses regulations related to well casing and drillings.  The language utilized is almost identical to that in the SDWA.  New York’s UIC program is similar, if not even stricter, than the federal UIC program regulations.

The study also criticizes the industry’s CWA exemption as a reason for pollution.  The CWA regulates water discharges under a program called the National Pollution Discharge Eliminations System (NPDES).  States have the option to utilize the federal program or adopt their own state program under more stringent standards.  New York State adopted its own program which is called the State Pollution Discharge Elimination System (SPDES) program.  This program permits and regulates how wastewater in New York State is disposed of, including wastewater from high volume hydraulic fracturing (HVHF) and natural gas development activities. The New York State Department of Environmental Conservation (DEC) went a step further and is in the process of developing special SPDES regulations specifically for HVHF.  These regulations are awaiting finalization and are specific to wastewater treatment plants that anticipate handling wastewater from natural gas development operations.  Furthermore, the only “exemption” for natural gas activities under the CWA deal with stormwater management plans and many states, including New York, have filled this void with more stringent regulations than what the CWA provides.

Some Very Rich Irony – Cornell University’s New Combined Heat and Power Plant Which Employs Natural Gas and Solar Power to Meet Needs of Researchers Such As Elaine Hill, Who Attack Gas

When discussing water impacts, Hill uses the term “cocktail of chemicals,” a derisive term term employed by anti-natural gas interests and another tell-tale sign of her ideological agenda.  She also refers to Theo Colburn’s work.  Colburn, an outspoken critic of natural gas development, has previously ginned up supposed issues regarding the potential health effects of chemicals utilized in the fracturing process.  Some of the chemicals she raises the alarm about have, of course, been used in other industrial and domestic processes without any widespread negative consequences.

Understandably, concerns with endocrine disruptors among any of the chemicals utilized should not be minimized.  Nevertheless, as with any other process that uses chemicals, there are protective measures in place to prevent exposure.  The industry works to protect water aquifers by mitigating potential risks, through the utilization of Best Management Practices.  These practices have been developed and standardized by the American Petroleum Industry and can be found here, here and here.  These guidelines include mitigation strategies to minimize surface impacts, well design and integrity guidelines, and water management strategies.

Moreover, just because the chemicals are present does not mean people will be affected.  The chemicals (like the gasoline in your car’s tank) are contained and separated from exposure routes.  Should there be a situation where exposure can happen (human or mechanical failure) these already developed best practices are designed to minimize any impact. Many are found in regulations and others are simply a matter of industry practice.

Air emissions related to natural gas development, for instance, create some exposure potential.  Desire to mitigate these concerns resulted, on April 17, 2012, in the the federal government passing the Oil and Natural Gas Air Pollution Standards. These standards, which are contained in 588 pages of regulations, are important when it comes to protecting both human health and infant health in particular.  These air standards are developed based on potential health effects assessed by understanding the potential exposure.

Exposure is the what (chemical), the how long (exposure length), the where (distance from source) and the how much (concentration). If the exposure levels are high enough and occur at a long enough time frame, then a health effect is observed. Regulations are developed to ensure that health effects do not occur. Regulators and industry keep track of air exposure data by tracking concentrations of air emissions from various sources and by also monitoring the spatial patterns of chemicals as they move through the air.  The EPA, to address cumulative effects of pollutants, has developed an emission inventory program that monitors trends in air emissions.  Air quality standards can be derived based on the inventory, available technology and health standards.

There are two good examples of air monitoring programs instituted to monitor air quality around shale development.  One is the Zielinska at al. 2010 Barnett Shale Study – Air Monitoring.  The objective of this study “was to provide a better understanding of the potential contributions of emission from gas production operations to population exposure to air toxics.”  The paper characterized the chemicals composition of the emissions and estimated the potential impact by measuring pollutants at different points from the source, creating a pollutant gradient.  The researchers then determined the concentrations of the selected air toxics within the communities in the Barnett Shale.  While the results were preliminary and the authors suggested conclusions should not be drawn from the study, results showed pollutant concentrations were generally low in the communities surveyed (less than 1 ppb).

The Pennsylvania Department of Environmental Protection (DEP) has also conducted short-term air monitoring studies to monitor the impact of natural gas activities.

• Northeastern Pennsylvania Marcellus Shale – Short-Term Ambient Air Sampling Report: “Results of the limited ambient air sampling initiative in the northeast region did not identify concentrations of any compound that would likely trigger air related health issues associated with Marcellus Shale drilling activities.”
• Northcentral Pennsylvania Marcellus Shale – Short-Term Ambient Air Sampling Report: “Results of the limited ambient air sampling initiative in the north central region did not identify concentrations of any compound that would likely trigger air-related health issues associated with Marcellus Shale drilling activities.”
• Southwestern Pennsylvania Marcellus Shale – Short-Term Ambient Air Sampling Report: “Results of the limited ambient air sampling initiative conducted in the southwest region did not identify concentrations of any compound that would likely trigger air-related health issues associated with Marcellus Shale drilling activities.”

While certain chemicals were detected in the air, they were determined to be below threshold values that would have caused a health effect.  So, where is that pathway?

Confounding Confirmation Bias

There are many confounding factors that were not considered in this study: lifestyle choices, maternal health history, familial genetics, occupational effects, randomness and other types of unidentified factors. These factors could easily have had an impact on the conclusion and should have been addressed.

As the discussion on the development of natural gas moves forward, it is expected many more studies will come forward. Unfortunately, the true nature of the issues is being lost in conformation bias.  Too many researchers and people reading their studies are only utilizing the evidence and information that supports their beliefs, while rejecting all other facts. This tendency jades research and objectivity resulting in questionable validity.

It is obvious Elaine Hill does not understand disease progression, pollution gradients and exposure pathways.  A statistical causal relationship between infant health and natural gas development cannot be confirmed with her research because the omission of important confounders. The results would have been more plausible if an exposure pathway analysis had been conducted.  However, because there is no pathway linking the two, such a link cannot be confirmed.  It is not likely to be confirmed either, given the variability of the results with lower increases in LBW closer to natural gas development than further away.  Add to the this the fact birth weights have been declining across the U.S. and what is left of Ms. Hill’s argument?