Starting the Analysis

In order to conduct a thorough analysis of how international trade affects pollution across the world, I use two packages - pandas for data organization, and plotly for visualization.

import warnings
warnings.filterwarnings("ignore", message="A NumPy version >=")

# Backbone of data organization
import pandas as pd
# Necessary for visualizations
import plotly.express as px

%matplotlib inline

Initial Exploration

I start with the University of Gothenberg's Quality of Government (QoG) Institute's Basic Dataset (2024). To achieve a rudimentary understanding of the relationship between international trade and pollution, I load the dataset and locate variables that could be used as operational definitions.

basecross = pd.read_csv('qog_bas_cs_jan24.csv')
basecross.head()

	ccode	cname	ccode_qog	cname_qog	ccodealp	ccodecow	version	ajr_settmort	atop_ally	atop_number	...	wvs_imprel	wvs_pmi12	wvs_psarmy	wvs_psdem	wvs_psexp	wvs_pssl	wvs_relacc	wvs_satfin	wvs_subh	wvs_trust
0	4	Afghanistan	4	Afghanistan	AFG	700.0	QoGBasCSjan24	4.540098	1.0	1.0	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
1	8	Albania	8	Albania	ALB	339.0	QoGBasCSjan24	NaN	1.0	8.0	...	2.869328	NaN	1.596485	3.849031	3.475513	1.744196	NaN	NaN	3.488758	0.027857
2	12	Algeria	12	Algeria	DZA	615.0	QoGBasCSjan24	4.359270	1.0	9.0	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
3	20	Andorra	20	Andorra	AND	232.0	QoGBasCSjan24	NaN	1.0	2.0	...	2.034930	2.710393	1.336049	3.681363	2.635721	1.830491	1.751004	6.561316	4.089642	0.255744
4	24	Angola	24	Angola	AGO	540.0	QoGBasCSjan24	5.634790	1.0	8.0	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN

5 rows × 338 columns

Variables of Interest

From the QOG Basic Dataset 2024 codebook, multiple variables pertaining to the relationship between international trade and pollution can be ascertained. Two such variables may be used here; international trade may be operationally defined by the variable for economic globalization (dr_eg), which ranks countries on a scale of 1-100 based on the flow of goods and services to other countries. In turn, pollution can be defined by the variable for the Environmental Performance Index (EPI) score (epi_epi), which ranks countries on a scale of 0-100 based on 32 different metrics of environmental health.

basecross.dr_eg

0      28.830755
1      63.483410
2      33.879074
3      70.329048
4      44.303589
         ...    
189    51.683723
190    27.821911
191    46.810032
192    40.550980
193    59.640228
Name: dr_eg, Length: 194, dtype: float64

basecross.epi_epi

0      43.599998
1      47.099998
2      29.600000
3            NaN
4      30.500000
         ...    
189    38.200001
190    46.400002
191    36.400002
192          NaN
193    38.400002
Name: epi_epi, Length: 194, dtype: float64

fig1 = px.scatter(
    data_frame = basecross,
    x = 'dr_eg',
    y = 'epi_epi',
    title = "Economic Globalization vs EPI Score by Country",
    labels={  
        'dr_eg': 'Economic Globalization',
        'epi_epi': 'EPI Score'
        },
    trendline = 'ols',
    hover_data = ['cname']
)
fig1

Initial Findings and Re-Testing

Initial findings from the basic dataset show a positive correlation between economic globalization and EPI score in which the greater a country's degree of economic globalization, the greater their performance on the EPI. This suggests that the more international trade a country conducts, the less polluted it is.

However, this seems somewhat counter-intuitive owing to recent research on this connection. According to the Grantham Research Institute on Climate Change and the Environment (2023), international trade is responsible for nearly 30% of global CO2 emissions, owing to the environmental cost of transporting goods across borders.

How could it be, then, that we are witnessing a positive correlation? It may be that I chose a poor metric for trade in economic globalization. For instance, the portion that a country's trade contributes to their gdp (wdi_trade) would serve as a more concrete definition for international trade. Comparing the outputs of using different variables side by side is possible with a scatterplot matrix.

basecross.wdi_trade

0            NaN
1      59.829731
2      45.330509
3            NaN
4      55.375816
         ...    
189    61.839191
190          NaN
191    77.483597
192    49.303493
193    79.325485
Name: wdi_trade, Length: 194, dtype: float64

fig2 = px.scatter_matrix(
    data_frame = basecross, 
    dimensions = ['dr_eg', 'wdi_trade', 'epi_epi'],
    title = "EPI Score vs Economic Globalization vs Trade % of GDP by Country",
    labels = {  
        'dr_eg':'Economic Globalization',
        'wdi_trade':'Trade % of GDP',
        'epi_epi':'EPI Score'
        },
    template = 'seaborn',
    hover_data = 'cname'
)
fig2.update_traces(diagonal_visible = False)
fig2.update_layout(width = 700, height = 700)
fig2

Creating More Specific Analyses

Using the scatterplot matrix, we can view the intersections of our three variables of interest. Unlike how prior research had demonstrated, EPI Score seemed to be positively correlated with both Economic Globalization and the Trade % of GDP. It may be, then, that the other items that comprise a country's EPI score outweigh the negative contribution of international trade on atmospheric pollution. In this sense, international trade may benefit the environment in certain countries except for when it comes to greenhouse emissions.

To find out, and observe the role of time in our analyses, I can take the time-series version of the QOG's Basic Dataset (2024), which takes critical observations of countries throughout the years, and the QOG's Environmental Indicators Dataset (2023), which has variables that allow for more in-depth analyses of pollution.

basetime = pd.read_csv('qog_bas_ts_jan24.csv')
basetime.head(5)

	ccode	cname	year	ccode_qog	cname_qog	ccodealp	ccodecow	version	cname_year	ccodealp_year	...	wdi_trade	wdi_unempfilo	wdi_unempilo	wdi_unempmilo	wdi_unempyfilo	wdi_unempyilo	wdi_unempymilo	wdi_wip	who_sanittot	whr_hap
0	4	Afghanistan	1946	4	Afghanistan	AFG	700.0	QoGBasTSjan24	Afghanistan 1946	AFG46	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
1	4	Afghanistan	1947	4	Afghanistan	AFG	700.0	QoGBasTSjan24	Afghanistan 1947	AFG47	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
2	4	Afghanistan	1948	4	Afghanistan	AFG	700.0	QoGBasTSjan24	Afghanistan 1948	AFG48	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
3	4	Afghanistan	1949	4	Afghanistan	AFG	700.0	QoGBasTSjan24	Afghanistan 1949	AFG49	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
4	4	Afghanistan	1950	4	Afghanistan	AFG	700.0	QoGBasTSjan24	Afghanistan 1950	AFG50	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN

5 rows × 252 columns

enviro = pd.read_csv('qog_ei_ts_sept21.csv')
enviro.head(5)

	Unnamed: 0	cname	ccode	year	cname_qog	ccode_qog	ccodealp	ccodealp_year	ccodecow	ccodevdem	...	wdi_precip	wdi_tpa	wvs_ameop	wvs_ceom	wvs_deop	wvs_epmip	wvs_epmpp	wvs_imeop	wvs_pedp	wvs_ploem
0	1	Afghanistan	4.0	1946	Afghanistan	4	AFG	AFG46	700.0	36.0	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
1	2	Afghanistan	4.0	1947	Afghanistan	4	AFG	AFG47	700.0	36.0	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
2	3	Afghanistan	4.0	1948	Afghanistan	4	AFG	AFG48	700.0	36.0	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
3	4	Afghanistan	4.0	1949	Afghanistan	4	AFG	AFG49	700.0	36.0	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
4	5	Afghanistan	4.0	1950	Afghanistan	4	AFG	AFG50	700.0	36.0	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN

5 rows × 415 columns

Merging Datasets

From a brief analysis of both datasets, it can be seen that there are commonalities in the data that can be used to perform an outer join. I can join specifically by country name (cname) and year of observation (year).

result = pd.merge(basetime, enviro, how = 'outer', on = ['cname', 'year'])
result.head(5)

	ccode_x	cname	year	ccode_qog_x	cname_qog_x	ccodealp_x	ccodecow_x	version_x	cname_year_x	ccodealp_year_x	...	wdi_precip	wdi_tpa	wvs_ameop	wvs_ceom	wvs_deop	wvs_epmip	wvs_epmpp	wvs_imeop	wvs_pedp	wvs_ploem
0	4.0	Afghanistan	1946	4.0	Afghanistan	AFG	700.0	QoGBasTSjan24	Afghanistan 1946	AFG46	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
1	4.0	Afghanistan	1947	4.0	Afghanistan	AFG	700.0	QoGBasTSjan24	Afghanistan 1947	AFG47	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
2	4.0	Afghanistan	1948	4.0	Afghanistan	AFG	700.0	QoGBasTSjan24	Afghanistan 1948	AFG48	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
3	4.0	Afghanistan	1949	4.0	Afghanistan	AFG	700.0	QoGBasTSjan24	Afghanistan 1949	AFG49	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
4	4.0	Afghanistan	1950	4.0	Afghanistan	AFG	700.0	QoGBasTSjan24	Afghanistan 1950	AFG50	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN

5 rows × 665 columns

International Trade vs CO2 emissions

Now that the datasets are merged, numerous other comparisons can be made using the various metrics of pollution within the environmental dataset. To specifically evalute the impact of international trade on CO2 emissions, I can use a country's total CO2 emissions in kilotons (edgar_co2t) and compare it with the proportion of GDP produced by trade. Additionally, now that I have longitudinal data, I can conduct this same analysis for all years between 2000 and now.

fig3 = px.scatter( 
    data_frame = result.query('year >= 2000 & year <=2023'),
    x = 'wdi_trade',
    y = 'edgar_co2t',
    animation_frame = "year", 
    animation_group = "cname", 
    title = "Trade % of GDP vs Total CO2 emissions (kt) from 1995-2020",
        labels={  
        'wdi_trade': 'Trade % of GDP',
        'edgar_co2t': 'Total CO2 emissions (kt)'
        },
    trendline = 'ols',
    hover_data = ['cname']
)
fig3

Analyzing International Trade vs CO2 emissions

Despite recent research showing that international trade is responsible for 30% of global CO2 emissions, it seems as though there is not much of a relationship. Given that the least squares line stays mostly horizontal, much like the data for each country throughout the years, it is difficult to say there is much correlation. Most countries appear to keep consistent in both their trading and yearly CO2 emissions, save for some outliers like China. Even zooming into the bottom half of the graph where most of the data is located does not yield much in terms of possible correlation. Therefore, this graph illustrates a lack of connection between international trade and pollution in terms of CO2 emissions.

International Trade vs Drinking Water Quality

Another important metric of pollution is water scarcity. According to Zhong et al. (2023), the degree of international trade can actually reduce water scarcity for higher-income countries, but exacerbate it for lower-income countries. This is perhaps due to the role of water in manufacturing, the way in which factories pollute water, and how transportation by ship disturbs water. I can assign a color value to countries' GDP (gle_cgdpc) to test this phenomenon.

fig4 = px.scatter( 
    data_frame = result.query('year >= 2000 & year <=2023'),
    x = 'wdi_trade',
    y = 'epi_uwd',
    animation_frame = "year",
    animation_group = "cname",
    title = "Economic Globalization vs Drinking Water Quality by Country and GDP (1995-2020)",
        labels={  
        'wdi_trade': 'Trade % of GDP',
        'epi_uwd': 'Drinking Water Quality'
        },
    trendline = 'ols',
    color = 'gle_cgdpc',
    hover_data = ['cname']
)
fig4

Analyzing International Trade vs Drinking Water Quality

In the years for which GDP was observed, and thus color could be assigned to the graph, it is clearly visible that lower-income countries had worse drinking water quality than higher-income countries, but the relationship between water quality and international trade is more nebulous. Although the least squares line indicates a positive correlation, the points themselves do not align in a discernable pattern to suggest as much. An interesting outlier in this regard is Qatar, which seems to have the highest drinking water quality of all despite being its middling GDP and trade.

It would also be valuable to identify this relationship in terms of imports and exports. Differentiating between a country's total sum of imports (gle_imp) and exports (gle_exp) in millions of dollars can illustrate the dynamic of water pollution between countries who produce goods for the international market and countries who buy them. However, I can only compare these statistics for the year 2000, given that total import and export were only observed then.

fig5 = px.scatter( 
    data_frame = result.query('year == 2000'),
    x = 'gle_imp',
    y = 'epi_uwd',
    title = "Total Import vs Drinking Water Quality by country (2000)",
        labels={  
        'gle_imp': 'Total Import (Millions of USD)',
        'epi_uwd': 'Drinking Water Quality'
        },
    trendline = 'ols',
    color = 'gle_cgdpc',
    hover_data = ['cname']
)
fig5

fig6 = px.scatter( 
    data_frame = result.query('year == 2000'),
    x = 'gle_exp',
    y = 'epi_uwd',
    title = "Total Export vs Drinking Water Quality by country (2000)",
        labels={  
        'gle_exp': 'Total Export (Millions of USD)',
        'epi_uwd': 'Drinking Water Quality'
        },
    trendline = 'ols',
    color = 'gle_cgdpc',
    hover_data = ['cname']
)
fig6

Conclusion

In conclusion, it would appear that international trade is overall negatively correlated with pollution. However, when it comes to CO2 emissions and drinking water quality specifically, the relationship is more nebulous. In order to discern which factors of pollution international trade benefits and which it exacerbates, in-depth analysis is requried for all 32 items of the EPI in relation to countries' capacity for international trade. This will be a necessary step in determining what actions need to be taken in order to most efficiiently minimize the environmental toll caused by the global market.

Dahlberg, Stefan, Aksel Sundström, Sören Holmberg, Bo Rothstein, Natalia Alvarado Pachon, Cem Mert Dalli, Rafael Lopez Valverde & Paula Nilsson. 2024. The Quality of Government Basic Dataset, version Jan24. University of Gothenburg: The Quality of Government Institute, https://www.gu.se/en/quality-government doi:10.18157/qogbasjan24

Grantham Research Institute on Climate Change and the Environment. 2023, June 12. How does trade contribute to climate change and how can it advance climate action?. London School of Economics and Political Science. https://www.lse.ac.uk/granthaminstitute/explainers/how-does-trade-contribute-to-climate-change-and-how-can-it-advance-climate-action/

Povitkina, Marina, Natalia Alvarado Pachon & Cem Mert Dalli. 2021. The Quality of Government Environmental Indicators Dataset, version Sep21. University of Gothenburg: The Quality of Government Institute, https://www.gu.se/en/quality-government

Zhong, R., Chen, A., Zhao, D., Mao, G., Zhao, X., Huang, H., & Liu, J. (2023). Impact of international trade on water scarcity: An assessment by improving the Falkenmark indicator. Journal of Cleaner Production, 385, 135740. https://doi.org/10.1016/j.jclepro.2022.135740

!python --version

Python 3.9.5