Climate Change

Students: Kristian Dolghier, Scott White, Benjamin Lentz, Ashley Marrs

Professor Alvin Murphy

December 13 2019

Introduction

Our goal is simple, create a regression of global temperature with regards to a few factors including Carbon Dioxide, Methane, and Nitrous Oxide concentrations in the atmosphere. While we cannot test for things such as hydrofluorocarbons and the amount of ozone in the atmosphere, this will show whether or not a percent change in Carbon Dioxide concentration will cause a temperature increase. With the current set of data we are using, we will only be conducting the regression for the past forty years as data from beforehand is incomplete. Whether or not our analysis will match the conclusion of the scientific community we will provide an analysis of the data from an impartial perspective and only state what we can see in the data.

Data Description:

We obtained the data used in this analysis from two different sources. From the World Bank, we found the global emissions of Carbon Dioxide (CO2), Nitrous Oxide (N2O) and Methane (CH4). Data on temperature was retrieved from NASA's GISS Surface Temperature Analysis report. Nitrous Oxide and Methane were measured from 1970 to 2012, while Carbon Dioxide and temperature were measured from 1960 to 2014. Due to this discrepancy, we dropped the observations outside of the 1970 to 2012 time range. See Table 1 for the summary statistics of each of the variables.

Table 1: Summary Statistics

Methodology & Modeling

We started by regressing temperature on CO2, N2O, and CH4. The coefficients were quite small, so we decided to attempt a model using log of CO2, CH4, and N2O. When we ran this model, the p-value on CH4 was 0.59 which showed us that it wasn't statistically significant so we removed it from the model. We then attempted to use a lagged model, running multiple regressions with multiple combinations of one or two lags on both CO2 and N2O. In all of these models, the p-values on the lagged coefficients exceeded .05 significantly, suggesting they are statistically insignificant. Additionally, the p-values of the coefficients on our non-lagged variables also became well over .05, suggesting their insignificance as well. Thus, we decided not to include lags in our model. Finally, we tested for heteroscedasticity using the Breusch-Pagan test and obtained a p-value of .7363, which is far greater than .05 and thus we accept the null hypothesis, H0, that there is constant variance, or homoscedasticity and there was no need to run a robust regression. As such, we decided on the linear-log model regressing CO2, and N2O on temperature.

Econometric Model

Table 2: Variable Descriptions

The model we have chosen is displayed below. Table 2 provides descriptions for each of the variables used in this models.

Results

The results of our chosen regression are displayed below. Listed below the equation are the standard errors in parentheses, the number of observations, the original R2 and adjusted R2.

Based on these results, we see that a one percent change in Kiloton of Carbon Dioxide emissions is associated with a .007593 degrees Fahrenheit increase in annual global temperature, all else constant. Additionally, a one percent change in Kiloton of nitrous oxide is associated with a .00435 degrees Fahrenheit increase in annual global average temperature, all else constant. To put this into perspective, the total global carbon emissions in the year of 2017 are 37,077 metric tons according to the European Commission. A one percent increase of this is 370.77 metric tons which is roughly the carbon output of Australia. This means that to increase global temperatures by 1 degree Fahrenheit, we must increase carbon emissions by 48,830 metric tons which is currently higher than the global output in one year. For one degree Celsius it is approximately 87894 metric tons. The R2 and adjusted R2 are quite similar, with only a small difference between the two. Considering the adjusted R2, we can see that this regression accounts for about 84% of the variation in annual global temperature. From this data we can conclude that carbon dioxide and nitrous oxide emissions do affect climate change. However, with our current growth rate of carbon emissions, we will see very slow growth in temperature not a sudden apocalyptic event.

Potential Limitations

A crucial limitation to our regression model is having a small sample size. Since planet Earth goes through natural climate cycles over tens of thousands of years, and the data we could find on Carbon Dioxide, Methane, and Nitrous Oxide only went to 1970, we might not have captured the isolated effect of each variable, independent of the natural climate cycle, on change in temperature. To mitigate this limitation, the data that we used on change in temperature was calculated by taking the difference in actual temperature and the predicted value using some estimator of the current natural climate cycle. Although we assume the predicted temperature values are accurate, we do not know the exact methodology behind the estimates.

Additionally, there are a few other variables which could be affecting annual global temperature change which we did not include in our model, such as hydrofluorocarbons, chloroflurocarbons and water vapor. We chose not to include these variables either because we deemed them to be irrelevant, or we were unable to find sufficient data, as was the case with water vapors. According to the EPA, global greenhouse gas emissions are divided into the following four categories, 76% Carbon Dioxide, 16% Methane, 6% Nitrous Oxide, and 2% F-gases which include Hydrofluorocarbons (HFCs), Perfluorocarbons (PFCs), and Sulfur Hexafluoride (SF6). An important note is that this only includes greenhouse global emission which does not consider things such as water vapor a greenhouse gas. The science for whether or not water is a greenhouse gas is still ongoing and we could not find any relevant data. Baring water vapor and F-gases, we have a majority of the currently defined greenhouse gases in our regression. However, this exclusion could mean our model suffers from omitted variable bias and the coefficients on CO2 and N2O may be biased to include the effects of these factors.

References

GISTEMP Team (2019). GISS Surface Temperature Analysis (GISTEMP), version 4. NASA Goddard Institute for Space Studies. Dataset accessed 2019-12-09 at https://data.giss.nasa.gov/gistemp/.

Lenssen, N., G. Schmidt, J. Hansen, M. Menne,A. Persin,R. Ruedy, and D. Zyss (2019). Improvements in the GISTEMP uncertainty model. J. Geophys. Res. Atmos., 124, no. 12, 6307-6326, doi:10.1029/2018JD029522.

World Bank (2012). Methane emissions (kt of CO2 equivalent) [Data file]. Retrieved from https://data.worldbank.org/indicator/EN.ATM.METH.KT.CE.

World Bank (2012). Nitrous oxide emissions (thousand metric tons of CO2 equivalent) [Data file]. Retrieved from https://data.worldbank.org/indicator/EN.ATM.NOXE.KT.CE.

World Bank (2014). CO2 emissions (kt) [Data file]. Retrieved from https://data.worldbank.org/indicator/EN.ATM.CO2E.KT.

"Energy, Climate Change, Environment." European Commission - European Commission, 12 Dec. 2019, ec.europa.eu/info/energy-climate-change-environment_en.