@article{TEXTUAL,
      recid = {1927},
      author = {Semprini, Jason},
      title = {Does exposure to wind turbines affect cancer incidence? A  quasi-experimental analysis linking SEER and Geological  Survey data in a hierarchical framework.},
      address = {2019-06-01},
      number = {TEXTUAL},
      abstract = {Objective: To conduct a series of multi-level regression  models to identify any association between wind turbines  and cancer incidence between counties, within counties over  time, and between cancer types.  Background: In a recent  campaign speech, President Trump declared that the noise  from wind turbines may cause cancer. This statement  resulted in tremendous outcry from political opponents and  wind energy advocates. But this question does not lack  merit from a scientific perspective. In fact, public health  researchers have begun to use large-scale population data  to evaluate the effect of wind turbine noise on a variety  of health issues. More so, biological studies have  identified noise as a potential mechanism driving certain  subtypes of breast cancer. Significance: Largely due to  confounding political factors and a lack of comprehensive  data on wind turbine development, no previous research has  directly investigated the effect of long-term exposure to  wind turbines on cancer incidence. However, multi-level  model analyses have been shown to mitigate bias against  possible inference by parsing out inherent county and  state-level characteristics. This framework can be utilized  to remove endogenous factors related to cancer and wind  turbine development. A hierarchical analysis linking SEER  data on cancer incidence with a newly released wind turbine  database creates an opportunity for econometric research to  further inform public health policy in the highly partisan  energy sector. Methods: Age-adjusted cancer incidence data  were obtained from the Surveillance Epidemiology End  Results Program (SEER). Wind turbine count and  characteristic data were accessed from the U.S. Wind  Turbine Database created by the U.S. Geological Survey  Program. In all, four hierarchical models were constructed:  one cross-sectional study nesting counties within states;  one three-level longitudinal model nesting repeated  measures within counties, within states; another  longitudinal model nesting repeated measures within  counties; and a final cross clustered model nesting  repeated measures within counties and cancer sites.  Independent variables included county and state-level  demographic controls associated with cancer incidence, as  well as wind turbine count and capacity measures. To  estimate the quasi-intervention effect of wind turbine  development at the county-level, each longitudinal model  incorporated a Difference-in-Difference analysis (DID).  Results: The cross-sectional analysis revealed that  counties with wind turbines had higher cancer incidence  than counties without (FE = 4.3, p = .06). However, for  counties with turbines, as exposure to higher quantities  per square mile increased, cancer incidence decreased (FE =  -26.7, p = .09). In each longitudinal analysis, the effect  of exposure completely disappeared. Exposure to higher  burden of wind turbines was shown to increase cancer  incidence, but this effect was marginal and only found in  one test (FE = 3.1, p = .06). Each of the  quasi-experimental designs brought similar results, that  the effect of early implementation of a wind turbines  program was protective against rising cancer incidence, but  with the onset of later development, the overall effect  became marginally significant (DID = -16., p > .5).  Finally, stratifying by cancer site revealed breast  cancer’s significant fixed and random effects from the  natural experiment (FE = 20.7, p = .002). Conclusion:  Greater exposure to wind turbines does not appear to  increase cancer incidence. And while the effect of turbine  development on cancer incidence varies in non-geographic  ways, possibly by cancer type, these findings remain  limited by the dearth of biological evidence suggesting any  causal effect from wind turbine exposure. Implications:  Without concrete, consensus support from the fields of  basic science and epidemiology, health policy analysis  lacks any ability to confront assumptions needed for  confident natural-experimentation. The causes of cancer  heterogeneity will remain a critical public health concern  until political platforms no longer hold greater legitimacy  than scientific investigation and public health priorities.  Acknowledgments: This research was part of the graduate  course, "Applied Hierarchical Linear Models" taught by Dr.  Jean Marshall. All analyses were completed using HLM7.  Access to this beta statistical software was generously  provided by Dr. Steve Raudenbush. A special thanks to Dr.  Marshall for the support throughout this project.},
      url = {http://knowledge.uchicago.edu/record/1927},
      doi = {https://doi.org/10.6082/uchicago.1927},
}