Panel data analysis in the demographic and spatial econometric estimation of carbon dioxide emissions sources, 1960-2010.

摘要

This dissertation presents three interrelated investigations focusing on the human dynamics of the carbon dioxide (CO2) emissions. Recent research has used the STIRPAT analytical framework for identifying anthropogenic sources of these emissions by modeling changes of population, affluence, and technology at a range of scales. Despite its wide applicability and flexibility, the STIRPAT framework as currently applied has several shortcomings with regards to modeling the spatial nature of economic production. The three investigations in this dissertation address this shortcoming by bringing space and geography in stochastic environmental modeling using concepts drawn from economic geography, quantitative spatial analysis, and economic demography.;The first investigation, in Chapter 3, addresses time and space effects in panel data. Exploring the consequences for ignoring divergence in undifferentiated time-series, cross-sectional data, this investigation illustrates potential problems for coefficients estimated using standard panel data procedures. Known as 'cluster confounding,' this effect results in the tendency for income to be positive over time, but negatively correlated with carbon dioxide between places, creating significant problems for estimation and inference in STIRPAT. I present a panel data regression technique for mitigating problems stemming from cluster confounding in panel data.;Chapter 4 examines the scale sensitivity hypothesis in STIRPAT, addressing long-standing criticisms of mathematical models in local-level analyses made within the literature of human and political ecology. Juxtaposing proximate physical sources of carbon dioxide emissions with distal 'theoretical' determinants, panel data estimates in this chapter illustrate weak support for the scale sensitivity hypothesis. By estimating labor force participation, age-structure, and retail employment as distal sources of CO2 emissions, and using industrial-economic base as proximate sources of CO2, this analysis challenges expected scale sensitivity hypotheses.;Last, Chapter 5 investigates the demographic dividend in national-level carbon dioxide emissions increases. Using panel data from 1960-2009, I test the temporal coincidence of industrial development and growth in labor force participation as an independent variable signaling dividend effects, and attempt to understand these interaction effects as drivers positively correlated with CO2 emissions. This analysis finds that demographic dividend effects are not statistically significant when strictly defined, and statistically significant when industrialization is more broadly defined.