Economics - Senior Research Proposal

...

In 2013, the National Research Council (NRC) uses the National Energy Modeling System to examine the effect of renewable energy tax provisions on greenhouse gas emissions in the U.S. The analysis compares the scenario where the PTC and ITC are effective through 2035 with a world without the provisions. The result indicates that an extension of the PTC and ITC through 2035 would lower CO2 emissions, but the impact is trivial, about 0.3 percent of US annual CO2 emissions over the projected time horizon (2012–2035). Although tax credits lead to an appreciable increase in renewable power generation, the total contribution of these sources is still small relative to the entire fleet of electricity generating units. The emissions effects therefore turn out to be small. Furthermore, when renewable portfolio standard (RPS), representing state mandates to promote renewable energy, is removed from the assumptions, the effects of the federal tax provisions on greenhouse gas emission roughly double, although they are still small relative to the economy’s emissions (0.5 percent). Another finding is that RPS has almost the same impact on mitigating CO2 emissions as the renewable electricity tax provisions, when each are examined separately (Murray et al., 2014).

Model

This paper will investigate the direct impact of the PTC and ITC on solar and wind production and greenhouse gas emission reduction using an empirical framework. There are two models to address two main questions. First, what is the role of the ITC and PTC in renewable energy production and investment in each state? What are other determinants of renewable energy production on the state level? The paper will construct a multiple linear regression model with cross-state panel data:

Ys,t = + ITCs, t–1 + PTCs, t–1 + Xs,t + + + es, t (1),[pic 1][pic 2][pic 3][pic 4][pic 5][pic 6]

where Ys, t is the dependent variable, measured by solar and wind investment/production by state and year; ITCs, t–1 and PTCs, t–1 present availability of the ITC and PTC (dummy variables: 1 = Yes), lagged by one year to control for policy response time; Xs, t is a vector that include other important state characteristics that potentially impact renewable energy production/investment; is state fixed effect to control for state-specific factors, for example: the recent North Dakota fracking boom may have negative impact on renewable investment only within North Dakota; is time fixed effect to control for abnormal time period such as the financial crisis in 2007 – 2009, which stunted economic activity in every industry; and es, t is the general error term. Control variables included in vector Xs,t are state’s per-capita income; residents’ green demand (percentage of residents who are members in environmental organizations); political belief (type of governor: liberal or conservative); level of education; state population; presence of other local policies to promote renewable energy (dummy variables: 1 = Yes); and locational effects (dummy variables for five regions: Northeast, Southeast, Northwest, West, Midwest). These independent variables have been mentioned and used by previous literatures (Price, 2002; Bird et al., 2005; Metcalf, 2007; Mendoca et al., 2009; Metcalf, 2010).[pic 7][pic 8]

The second question is a simple follow-up from the first model: How much does the increase in solar and wind capacity contribute to reducing carbon emission per capita, if at all? Following Abolhosseini et al. (2014), the paper will construct a cross-state panel data model:

s,t = + REGENP +s,t + + + es, t (2),[pic 9][pic 10][pic 11][pic 12][pic 13][pic 14]

where s,t is the dependent variable, measured by carbon emission caused by consumption of energy in metric ton per capita; REGENP presents share of solar and energy sources as a percentage of total power generated; s,t is a vector of other independent variables; + are error components specific to state and time; and the error term es, t. s,t might comprise of state income per capita, number of energy-related patents per capita, environmental tax per capita, locational effects, and time trend representing the rate of technical change in the CO2 function over time. According to Abolhosseini et al. (2014), the inclusion of trend squared and its interaction with explanatory variables allows testing for the nonlinearity and non-neutrality of the changes in the CO2 function. [pic 15][pic 16][pic 17][pic 18][pic 19]

Data

In model (1), the dependent variable is solar and wind investment and/or production measured in megawatts per state per year, which is available on Energy Information Administration (EIA) and Bloomberg New Energy Finance (BNEF). Depend on the statistical significance of the regression results, solar and wind may need to be assessed separately, as the PTC and ITC might impact two technologies in very different patterns. Independent variables include presence of ITC, PTC and other local incentive policies such as RPS and feed-in tariffs in form of dummy variables, which can be obtained on state government’s website, EIA, and BNEF. Regional effects are controlled for by five dummy variables for five regions: Northeast, Southeast, Northwest, West, and Midwest, which is categorized by the Consensus Bureau. Data for state income per capita, type of governor, level of citizen education, residents’ demand for green energy, and state population are widely available on the Consensus Bureau, Bureau of Economic Analysis, EIA, and other U.S. government websites.

In model (2), the dependent variable is carbon dioxide (CO2) emissions per capita generated by fossil fuels burning and the cement manufacturers, plus CO2 generated by consumption of solid, liquid and gas fuels, measured in metric ton per capita. State data is available on the Energy Protection Agency’s Emissions & Generation Resource Integrated Database (eGrid). The independent variable of interest is share of electricity produced by renewable energy sources in total power generation, reported by EIA and BNEF. Similar to model (1), locational effects are controlled for by five dummy variables for five regions. Explanatory variables such as state income per capita, number of energy-related patents per capita, and environmental tax are available on previously mentioned government portals.

A few variations of both models will be tested for different selection of independent variables and dependent variables to determine the best fit. More literature review and research are certainly needed to determine the appropriate measures of these explanatory variables, whether data is available on the state level for