Nexus of Economic Growth, Economic Structure, and Environmental Pollution: Using a Novel Machine Learning Approach

Abstract

The economy and environment still show complicated relationships, which have generated various and conflicting hypotheses. This study aims to propose a new perspective on the connection between economy and environment across 164 countries using an innovative clustering method, including Principal Components Analysis (PCA) and a machine learning approach. The outcome introduces three clusters of countries with similar economic and environmental characteristics. Cluster 1 constitutes countries with the highest levels of economic development and environmental quality. They include Luxembourg, Switzerland, Ireland, Norway, Singapore, the US, and Australia. Cluster 2 involves countries with less than the highest levels of economic development and environmental quality, covering the right side of the Environmental Kuznets Hypothesis (EKH) and the Pollution Halo Hypothesis (PHH-Halo). These include Qatar, Denmark, Iceland, The Netherlands, Austria, the UK, Germany, UAE, New Zealand, and Israel. Finally, the lowest development levels of economic and environmental development are apparent in the countries in Cluster 3, indicating the left side of the EKH and the Pollution Haven Hypothesis (PHH-Haven). This finding gathers the three hypotheses of EKH, PHH-Halo, and Haven in one unique framework of the economy–environment nexus.

1. Introduction

Environmental health is a critical challenge in the world due to serious environmental and ecological problems like pollution, climate change, and greenhouse gas emissions. Greenhouse gas emissions have increased global temperature and air pollution, which damage global ecosystems, increase sea levels, reduce polar ice volume, increase extreme weather, and damage health status [1,2,3]. These threats can be fatal to all forms of life, including human, animal, and vegetation, on Earth [4,5,6,7].

Environmental threats are mainly attributed to economic and industrial production, as well as transportation and inefficient management of waste [8,9,10]. They are rooted in different aspects of the economy, including fossil fuel, transportation, cement production, and the agricultural sector [11,12,13]. Likewise, the agricultural sector has exacerbated the danger of CO₂ emissions by damaging approximately one-fifth of forests, which are the main source of carbon sequestration and carbon sink on the Earth [14,15,16,17].

Despite their detrimental impacts, economic activities can ameliorate environmental degradation through technological advancement and enhancement of efficiency, which complicates the economy–environment relationship. Economic development is predominantly connected to technological advancement, with higher energy efficiency but lower pollutant by-products; for example, new LED lights, electric engines, and advanced lithium-ion batteries [18,19,20,21]. The wealth and income earned as a result of economic development allow economies to invest more in renewable energy production [22]. Correspondingly, the development of the construction industry can enhance building automation systems for further optimization of energy consumption in heating, lighting, and air-conditioning systems in buildings, leading to up to 30% reduction in energy use [23].

The Environmental Kuznets Hypothesis (EKH), Pollution Haven Hypothesis (PHH-Haven), and Pollution Halo Hypothesis (PHH-Halo) inclusively cover both the constructive and detrimental influences of economic ventures, while raising deeper complex ambiguities [24,25]. On the one hand, the EKH proposes that economic growth can damage the environment in developing economies while improving the environment in developed countries [26,27,28]. On the other hand, the PHH-Halo highlights the constructive impact of economic development through foreign investment on environmental development across developing countries [29,30,31,32]. However, the PHH-Haven shows how foreign investment pollutes the environment within developing economies [33,34,35,36].

Despite an extensive body of literature, there are still serious complexities and unaddressed issues regarding the EKH, PHH-Haven, and PHH-Halo. Despite the extensive exploration of these hypotheses, understanding the nexus of environment and economy remains elusive and the findings show inconsistencies. First, previous studies have separately examined each of these hypotheses in isolation, neglecting to investigate them all together in an integrated framework. This hinders mapping an all-inclusive image of the interactions between the economy and the environment. Second, the behavioral differences between various economic sectors are unexplored regarding the economy–environment nexus [37]. The previous studies have examined the relationships between economy and environment from an aggregated perspective or a specific sector, without sector-wide analysis allowing the comparison of various sectors [38]. Third, they have examined the EKH, PHH-Haven, and PHH-Halo mainly by using econometric approaches with linear regression modeling and quantitative representation, lacking clustering and categorization methodology with graphical reflection [39,40].

To address these issues, this study aims to examine the EKH and PHH-Haven at the global level to reveal from a clustering viewpoint how economic development is connected to environmental development. In addition, it innovatively checks the role of economic structure in this relationship. To this end, it uses a novel approach of machine learning and clustering to classify countries according to their environmental development, economic development, and economic structure. This novel approach can expand the boundaries of the EKH and PHH-Haven by interconnecting the traditional perspectives of these hypotheses with sector-wide analysis while employing the modern techniques of machine learning and clustering. Instead of a numerical and econometric approach, it can innovatively represent a visual and clustering reflection of the economy–environment nexus, opening a novel horizon to the literature on the relationship between economy and environment. Moreover, it can contribute to the literature on the usage of the term “nexus” and show whether it is a meaningful term revealing the real connections among the key drivers or a buzzword, disregarding the structural and institutional impacts.

2. Theoretical Framework

The EKH indicates two conflicting impacts of the economy on environmental pollution. Based on the hypothesis, the economy is beneficial in developed countries but harmful in developing economies [33,34,35,36]. Figure 1 demonstrates the non-linear connection of the economy with the environment, according to the EKH. The left side indicates their positive relationship within developing economies, which can be attributed to their intense concentration on economic growth rather than on conserving ecological and environmental quality. However, this direct nexus can be maintained until reaching a turning point. Then, the relationship transforms to a negative connection, which can indicate the conditions in developed countries. This shift is due to their utilization of efficient technologies requiring low energy and emitting reduced pollutant by-products [41,42,43].

The PHH-Haven and PHH-Halo describe two conflicting views on the impact of developed countries’ economic activities on the environment of developing countries. Figure 2 indicates the PHH-Haven and PHH-Halo. The left side, based on the PHH-Haven, represents the direct nexus of foreign investment and environmental pollution. Accordingly, foreign investment increases environmental pollution, which may be rooted in lax or no environmental tax in developing economies. Such weak environmental policy paves the way for increasingly attracting polluting investment from other countries [44,45,46].

However, the right side of the curve shows a decreasing trend, which shows a reverse relationship between foreign investment and environmental pollution according to the PHH-Halo. It posits that developed countries attract cleaner foreign investment, which could be due to their strong environmental regulations. Such environmental policies attract more efficient and environmentally friendly technologies [32,47].

These hypotheses, which reflect the economy–environment nexus, highlight “nexus” as a highly useful and meaningful term in the literature; for example, relating to water–energy–food, water–energy–migration, and environment–economy–social relationships. In this regard, Hussein and Ezbakhe (2022) confirmed the significance of this term, i.e., nexus, within the framework of Water–Employment–Migration (WEM) [48]. Similarly, Daher and Mohtar (2015) supported the considerable nexus of Water–Energy–Food (WEF) [49]. Analogously, Nodehi et al. (2022) revealed an Environment–Economy–Social nexus from an integrated sustainability perspective [50,51]. However, Allouche et al. (2015) criticized this approach by stating that nexus is a buzzword that masks structural, institutional, and political dynamics [52]. Therefore, an investigation of the nexus among economic and environmental drivers can address this research gap.

3. Methodology

Following [53], this research employs machine learning clustering to reveal the nexus of economy and environment. This method encompasses data processing, selection of the optimal number of clusters, conducting the clustering strategy, and Principal Component Analysis (PCA), conducted in this study using RStudio version 4.0.2 (R Core Team, 2020).

3.1. Data

The first step is the collection and processing of data, including the economic and environmental variables of 164 countries in 2020 from aggregated and disaggregated perspectives. The economic factors are Gross Domestic Product (GDP) per capita (constant 2015 USD), Services value-added per capita (constant 2015 USD), Industry (including construction) value-added per capita (constant 2015 USD), Manufacturing value-added per capita (constant 2015 USD), and Agriculture, forestry, and fishing value added per capita (constant 2015 USD). They are derived from the World Development Index database (World Bank, 2024). The environmental variables are Environmental Performance Index (EPI), Environmental Health, and Ecosystem Vitality, ranging from 0 to 100, indicating the most and least polluted cases, respectively, which are extracted from the Yale University Database.

The gathered data are normalized and categorized into two groups of aggregate and disaggregate. The aggregate class includes GDP and EPI. The economic variables of the disaggregate group include Services value-added per capita, Industry value-added per capita, Manufacturing value-added per capita, and Agriculture, forestry, and fishing value-added per capita. The economic variables of the disaggregate group include Environmental Health and Ecosystem Vitality. The normalization process uses a standard technique [54] according to Equation (1):

$$x_{std}={\displaystyle \frac{x_t-x_{min}}{x_{max}-x_{min}}}$$

(1)

where $x_{std}$ is the standardized value, $x_t$ is the current value, $x_{min}$ is the minimum value, and $x_{max}$ is the maximum value.

3.2. Optimal Number of Clusters

The second step is the selection of the number of clusters using the Calinski approach [54], represented in Equation (2):

$$C\left(k\right)={\displaystyle \frac{B_k}{W_k}}.{\displaystyle \frac{(n-k)}{(k-1)}}$$

(2)

where C(k) is the Calinski index with k number of clusters, B and W denote the between-cluster and within-cluster sum of squares, respectively, and n indicates the number of observations. A higher value for the Calinski index indicates a more accurate number of clusters.

3.3. Clustering

The study uses the K-means clustering method to categorize countries according to their level of economic development and environmental quality [55,56,57]. This approach minimizes the sum of the squared distances of values of the economic and environmental indexes with their mean values to categorize the sample into K clusters and define each one. To this end, it minimizes Formula (3).

$$\sum _{i=1}^k\sum _{x\in O_i}{(x-{\theta }_i)}^2$$

(3)

where $O_i$ denotes the observations of cluster I, and $\theta$ shows the centroid of the corresponding cluster. This process is repeated for the maximum iteration number or until a stabilized point is established. This technique classifies the sample countries, without overlapping, into the optimal number of clusters, according to the environmental and economic variables analyzed in the research.

In this research, the clustering method is superior to other alternative methods due to its scalability, non-learning framework, and innovativeness. This approach can accurately and efficiently analyze large samples, contrasting hierarchical clustering or density-based techniques, such as Density-Based Spatial Clustering of Applications with Noise (DBSCAN). In addition, this method investigates datasets without any presumptions or pre-learning steps, while alternative techniques, including Gaussian Mixture Models (GMM), establish weighting and probabilistic restrictions at the beginning of the estimations [55,56,57]. This feature enables the k-clustering method to mitigate potentially biased results. Moreover, this approach analyzes the nexus of economy–environment from a novel viewpoint, differing from the previous studies, which mainly employed econometric models.

3.4. Principal Components Analysis (PCA)

After clustering, the Principal Components Analysis (PCA) method projects the economic and environmental data onto the first two principal components to delineate the K clusters in two dimensions [58,59,60]. In this way, it indicates the economy–environment nexus among various countries of the world. Furthermore, this analysis geographically maps the clustered countries to indicate their locations. If countries in one cluster show high indexes both in economic and environmental indicators, this result confirms the right half of the EKH and the PHH-Halo [61,62]. If another cluster involves countries with low values for economic and environmental indexes, the result affirms the left half of the EKH and the PHH-Haven; otherwise, they support the rejection of the hypotheses. The employed methods delineate a comprehensive image of the connections between the economic and environmental characteristics within a correlative framework.

This PCA is more advantageous than the other analogous methods due to its high interpretability. It facilitates the interpretation of the clustering results by compacting multiple factors into a bi-dimensional framework. It further simplifies the interpretation of the results by representing them in a graphical structure. These advantages enable the PCA method to smooth the interpretation of the results without reducing their accuracy [63]. This is in contrast to the other non-linear methods, such as t-distributed Stochastic Neighbor Embedding (t-SNE) and Uniform Manifold Approximation and Projection (UMAP), because the increase in their interpretability can decrease their accuracies [53].

Furthermore, the employed methods concentrate on analyzing the EKH, PHH-Haven, and PHH-Halo from an associative perspective, rather than a causal framework. They focus principally on measuring the correlations between the economic and environmental factors to maximize the precision of the results. In this way, they can highlight the connections between the economic and environmental contributors to indicate if they follow a nonlinear pattern as in the EKH, PHH-Haven, and PHH-Halo. This analysis differs from conducting a causal analysis to check the direction of the relationship, i.e., whether the economy affects the environment or vice versa [64,65,66].

4. Results

The results categorize countries into three clusters: Cluster 1, including the richest countries with the highest environmental quality, Cluster 2 with high incomes and environmental quality, and Cluster 3 with the lowest incomes and environmental quality. According to this classification, more developed economies show higher quality environments, both at the aggregated and disaggregated analysis levels.

Table 1. Results of the Calinski method for an optimal number of clusters.

Number of Clusters	1	2	3#	4
Calinski Harabasz pseudo-F	-	447.29	542.94	374.75

Note: “# ” indicates the highest pseudo-F, that is, the optimal number of clusters.

represents the estimated pseudo-F statistics using the Calinski method. According to Table 1, the highest value of the estimated pseudo-F statistics is 542.94, which is indicated in the column with three clusters. Thus, the optimal clustering number in the sample dataset is three, which is applied in this study.

Figure 3 displays the results of categorizing the countries into three clusters, based on their economic and environmental scores, using clustering machine learning techniques. According to the figure, Dimension 1 on the horizontal axis explains 67.2% of variations in the aggregated and disaggregated environmental and economic indexes, while that of the y-axis is 22.2%, categorizing the countries into three clusters, including Cluster 1, Cluster 2, and Cluster 3, decoded by green, blue, and red colors, respectively.

Table A1. List of countries clustered, according to their economic and environmental scores.

		Environment			Economy
Country	Cluster	EPI	EH	EV	GDP	Serv.	Indu.	Manu.	Agri.
Luxembourg	1	82	93	75	104,616	83,563	11,480	5024	189
Switzerland	1	82	95	73	84,637	60,173	21,553	15,729	508
Ireland	1	73	94	59	79,442	41,954	31,353	28,855	699
Norway	1	78	99	64	75,287	41,891	23,967	4736	1119
Singapore	1	58	85	40	59,176	40,900	15,036	12,724	19
Qatar	2	37	57	24	58,476	27,881	33,310	5120	175
United States of America	1	69	83	60	58,452	45,004	10,782	6664	685
Australia	1	75	92	64	58,082	40,353	13,039	3193	999
Denmark	2	83	92	76	55,654	36,168	11,344	7608	494
Iceland	2	72	98	55	52,486	33,761	10,657	4901	2740
Sweden	2	79	98	66	51,953	33,991	11,311	6565	676
The Netherlands	2	75	91	65	46,303	32,033	8694	5242	787
Finland	2	79	99	65	44,985	27,269	10,584	6759	1064
Austria	2	80	88	74	43,344	27,028	11,264	7523	479
Canada	2	71	92	57	42,390	28,941	9969	3925	834
United Kingdom	2	81	92	74	42,192	29,601	8049	4358	246
Germany	2	77	90	69	41,602	25,945	11,064	8127	277
United Arab Emirates	2	56	55	56	41,276	23,497	17,549	4551	350
Belgium	2	73	86	65	40,656	28,359	7880	5044	245
New Zealand	2	71	88	60	39,608	26,488	8113	4114	1659
Israel	2	66	84	54	38,177	26,179	7364	4877	461
France	2	80	92	72	35,807	25,413	6074	3634	524
Japan	2	75	90	65	34,651	23,843	10,185	7190	316
South Korea	2	67	81	57	31,378	17,675	10,570	8251	536
Brunei Darussalam	2	55	74	42	30,402	11,712	18,907	6004	340
Italy	2	71	86	61	29,375	19,743	6252	4158	587
Cyprus	2	65	82	54	27,090	14,285	2520	1117	332
Spain	2	74	87	66	24,830	16,931	4931	2601	713
Kuwait	2	54	57	51	23,929	15,747	12,704	1624	133
Slovenia	3	72	69	74	22,964	12,771	6897	5043	485
Bahamas	2	44	53	37	22,831	19,574	2314	287	189
Bahrain	3	51	49	52	22,579	12,946	8934	3822	73
Estonia	3	65	73	60	20,118	12,456	4696	2588	485
Portugal	3	67	83	56	19,802	12,853	3953	2458	405
Czech Republic	3	71	68	73	19,048	10,731	6015	4516	476
Saudi Arabia	3	44	47	42	18,857	10,142	7762	2297	583
Oman	3	39	43	35	17,662	8560	9351	1540	499
Slovakia	3	68	64	71	17,617	10,484	4886	3322	356
Greece	3	69	81	61	17,283	11,687	2777	1692	704
Lithuania	3	63	63	63	17,248	10,304	4660	3034	566
Uruguay	3	49	68	37	16,460	9879	4043	2086	964
Seychelles	3	58	51	63	15,322	10,393	2102	957	413
Latvia	3	62	58	64	15,312	9680	3060	1773	563
Barbados	3	46	61	36	15,067	10,921	2170	852	286
Antigua and Barbuda	3	49	56	44	14,804	9857	3093	409	272
Poland	3	61	59	62	14,775	8504	4225	2562	327
Hungary	3	64	54	70	14,430	8104	3535	2588	485
Trinidad and Tobago	3	48	55	43	14,214	8657	4815	2006	242
Croatia	3	63	61	64	13,075	7646	2865	1684	418
Chile	3	55	63	50	12,739	7417	3505	1335	501
Panama	3	47	50	45	12,307	8875	2694	662	393
Turkey	3	43	51	37	12,180	6700	3262	2009	811
Costa Rica	3	53	61	47	12,030	8182	2411	1519	578
Argentina	3	52	60	47	11,347	6455	2449	1484	639
Kazakhstan	3	45	41	47	10,974	6193	3764	1255	553
Romania	3	65	50	74	10,899	6210	3081	1972	440
Malaysia	3	48	55	43	10,383	5705	3795	2372	764
Russia	3	51	53	49	9714	5555	2966	1324	399
Mauritius	3	45	60	35	9363	6274	1613	1051	304
Mexico	3	53	48	56	9274	5793	2683	1827	328
Guyana	3	36	34	38	9127	2422	4615	291	1512
Saint Lucia	3	43	48	40	8335	5965	1041	321	220
Brazil	3	51	50	52	8256	5177	1508	834	419
Saint Vincent and the Grenadines	3	48	44	51	7948	5089	1074	341	548
Grenada	3	43	46	41	7915	5203	1153	300	390
Dominican Republic	3	46	36	53	7572	4275	2172	1073	448
Suriname	3	45	37	51	7275	4199	1958	1238	548
Cuba	3	48	51	47	7173	5334	1568	857	217
Maldives	3	36	48	27	7164	5316	848	182	502
Gabon	3	46	28	58	6620	2665	2916	1142	420
Dominica	3	45	47	43	6566	3857	864	212	852
Serbia	3	55	48	60	6552	3316	1705	932	426
Montenegro	3	46	47	46	6516	3769	1159	277	541
Lebanon	3	45	53	40	6486	5109	1026	429	346
Belarus	3	53	56	51	6252	3031	1880	1346	374
Paraguay	3	46	47	46	6095	2986	2083	1180	590
Thailand	3	45	48	44	6048	3453	2064	1567	524
Equatorial Guinea	3	38	28	45	5996	2923	3155	140	140
Colombia	3	53	55	52	5853	3506	1396	669	399
Botswana	3	40	20	54	5811	3768	1784	347	117
Peru	3	44	45	43	5754	3184	1645	723	473
South Africa	3	43	31	51	5749	3840	1207	658	165
Bosnia and Herzegovina	3	45	44	46	5459	2993	1197	612	337
Ecuador	3	51	50	52	5332	2798	1582	781	580
Iran	3	48	48	48	5141	2806	1596	743	563
Azerbaijan	3	47	33	56	5084	2384	2056	333	397
North Macedonia	3	55	44	63	5067	2912	1067	582	432
Marshall Islands	3	31	33	30	5060	3478	545	82	1053
Belize	3	42	40	43	4958	3000	841	380	508
Jamaica	3	48	46	50	4770	2886	969	404	356
Fiji	3	34	35	34	4701	2473	810	526	442
Tonga	3	45	44	46	4638	2469	686	264	723
Georgia	3	41	39	43	4448	2814	817	368	349
Albania	3	49	45	52	4419	2101	962	288	843
Armenia	3	52	44	58	4256	2265	1174	503	493
Sri Lanka	3	39	42	37	4230	2407	1251	703	316
Samoa	3	37	42	34	4224	3097	511	176	322
Iraq	3	40	40	39	4170	2117	1832	107	339
Namibia	3	40	23	52	4152	2523	1050	486	341
Guatemala	3	32	31	33	4124	2545	921	594	405
Mongolia	3	32	28	35	4107	1871	1177	343	619
Algeria	3	45	50	41	3874	2016	1320	784	492
Egypt	3	43	34	50	3836	2007	1314	572	409
Jordan	3	53	59	50	3785	2300	948	685	172
Indonesia	3	38	29	44	3780	1699	1443	759	503
El Salvador	3	43	43	44	3776	2304	917	552	211
Tunisia	3	47	49	45	3699	2219	829	490	381
Eswatini	3	34	18	45	3673	1992	1221	1076	279
Viet Nam	3	33	41	29	3352	1425	1219	829	414
Philippines	3	38	34	41	3196	1931	948	619	317
Moldova	3	44	46	44	3189	1676	763	356	317
Uzbekistan	3	44	30	54	3187	1290	769	436	838
Morocco	3	42	33	48	3082	1649	826	460	322
Bolivia	3	44	36	50	2920	1403	669	310	365
Bhutan	3	39	30	46	2884	1360	980	176	437
Micronesia	3	33	31	35	2883	1894	152	19	662
Cabo Verde	3	33	30	35	2876	1937	354	143	177
Djibouti	3	28	20	33	2815	2207	399	115	44
Papua New Guinea	3	32	28	35	2455	1074	837	37	445
Angola	3	30	20	36	2435	1264	885	171	254
Nigeria	3	31	14	42	2401	1376	436	213	562
Ukraine	3	50	49	50	2350	1188	477	256	252
Cote d’Ivoire	3	26	19	30	2235	1121	495	310	452
Honduras	3	38	33	41	2191	1356	523	355	279
Solomon Islands	3	27	20	31	2080	1089	341	191	651
Timor-Leste	3	35	29	40	1987	775	995	24	221
Ghana	3	28	20	33	1951	737	660	227	394
Nicaragua	3	39	40	39	1904	909	465	277	366
India	3	28	16	35	1814	869	484	287	313
Republic of Congo	3	31	18	39	1741	642	856	172	147
Kenya	3	35	26	41	1617	936	299	146	299
Bangladesh	3	29	22	34	1593	837	479	295	202
Mauritania	3	28	20	33	1588	776	291	103	348
Pakistan	3	33	15	45	1578	848	296	183	343
Sao Tome and Principe	3	38	29	44	1423	999	160	16	144
Cameroon	3	34	14	47	1419	711	357	198	240
Cambodia	3	34	31	36	1404	532	453	227	305
Senegal	3	31	20	38	1391	700	312	222	229
Kiribati	3	38	23	48	1358	948	129	58	360
Haiti	3	27	22	31	1324	741	306	223	224
Myanmar	3	25	25	25	1295	541	464	281	291
Zambia	3	35	21	44	1237	688	431	97	63
Zimbabwe	3	37	23	47	1213	741	258	114	123
Benin	3	30	20	36	1165	556	187	114	320
Kyrgyzstan	3	40	34	44	1118	553	306	171	159
Tanzania	3	31	27	34	1029	379	289	84	252
Nepal	3	33	21	41	1018	512	140	50	256
Guinea	3	26	19	32	959	377	316	101	188
Lesotho	3	28	12	39	944	505	297	173	48
Uganda	3	36	26	42	918	404	261	153	209
Togo	3	30	16	38	830	405	207	140	155
Rwanda	3	34	24	40	822	410	149	65	190
Ethiopia	3	34	25	41	811	331	187	46	248
Mali	3	29	20	36	746	287	128	41	276
Burkina Faso	3	38	20	51	709	306	193	75	140
Gambia	3	28	21	32	656	338	134	19	130
Chad	3	27	15	35	622	177	214	17	228
Sierra Leone	3	26	19	30	608	197	28	10	367
Mozambique	3	34	28	38	584	257	109	45	145
Afghanistan	3	26	20	29	529	250	131	85	136
Niger	3	31	17	40	520	190	111	40	192
Dem. Rep. Congo	3	36	22	46	487	156	224	76	85
Madagascar	3	27	22	29	434	218	75	36	111
Central African Republic	3	37	12	54	375	175	82	73	40

Note: Green, blue, and red colors in the left (country) column show Clusters 1, 2, and 3 of countries, respectively. For the columns related to the environment performance index and economy sectors, the darkness of the green and red colors shows high and low degrees of indicators, respectively. EPI is Environmental Performance Index, measured between 1 and 100; EH denotes Environmental Health, measured between 1 and 100; EV signifies Ecosystem Vitality measured between 1 and 100; GDP indicates Gross Domestic Product per capita, measured in constant 2015 USD; Serv. shows Services value added per capita, measured in constant 2015 USD; Indu. denotes Industry (including construction) value-added per capita, measured in constant 2015 USD; Manu. denotes Manufacturing value added per capita, measured in constant 2015 USD; Agri. represents Agriculture, forestry, and fishing value added per capita, measured in constant 2015 USD.

in the Appendix A represents the details of the three clusters, the countries, the environmental indicators and sub-indicators, and the income of the whole economy and each sector, described below. It provides the GDP and sectoral value added per capita of the services, industrial, manufacturing, and agricultural sectors while clustering the countries.

4.1. Cluster 1 (Richest Countries)

Cluster 1 indicates the highest levels of income and environmental quality, with seven countries including Luxembourg, Switzerland, Ireland, Norway, Singapore, the US, and Australia. These countries have the five highest levels of income, including Luxembourg, Switzerland, Ireland, Norway, and Singapore per capita GDP ranging from USD 104,616 to 59,176. The other two countries of this cluster, the US and Australia, take the 7th and 8th places with USD 58,452 and 58,082, respectively. In addition to the total GDP, they show the highest added value in various economic sectors. In the services sector, all seven countries possess the highest per capita value added, ranging between 40,353 and 83,563. Although their ranks in the industrial and agricultural sectors are not as high as in the services sector, they still have high added values. In the industrial sector, Ireland, Norway, Switzerland, Singapore, Australia, Luxembourg, and the US occupy the 2nd, 3rd, 4th, 7th, 8th, and 15th ranks, respectively. In the agricultural sector, Ireland, Switzerland, and Singapore take the first three positions with USD 28,855, 15,729, and 12,724, respectively. The US, Luxembourg, Norway, and Australia occupy the 10th, 17th, 20th, and 30th ranks, respectively.

Similar to economic status, the highest environmental scores go to the countries in Cluster 1. The 2nd and 3rd highest EPI scores go to Luxembourg and Switzerland, with 82.3 and 81.5, respectively. Subsequently, Norway, Australia, Ireland, the US, and Singapore are in 9th, 13th, 16th, 23rd, and 38th places with EPI scores of 77.7, 74.9, 72.8, 69.3, and 58.1, respectively. Similarly, in terms of the Environmental Health index, they have the highest ranks. In this regard, Norway, Switzerland, Ireland, Luxembourg, Australia, Singapore, and the US are in the 2nd, 5th, 6th, 7th, 11th, 22nd, and 25th ranks with 98.5, 95.0, 94.0, 92.6, 91.6, 85.0, and 82.8 scores, respectively. Analogously, the scores for Ecosystem Vitality are high in this cluster. For example, they are 75.4, 72.5, 63.8, 63.8, 60.3, 60.2, and 58.6 in Luxembourg, Switzerland, Norway, Australia, the US, and Ireland, respectively, in the 2nd, 8th, 21st, 22nd, 29th, and 33rd ranks. Therefore, Cluster 1 has not only the richest countries but also the greatest environmental quality, which implies that countries with developed economies have high quality of environment, consistent with the right side of EKH and the PHH-Halo.

4.2. Cluster 2 (Rich Countries)

Cluster 2 shows a high score of environmental performance and income level, albeit below Cluster 1. These countries include Qatar, Denmark, Iceland, Sweden, The Netherlands, Finland, Austria, Canada, the United Kingdom, Germany, UAE, Belgium, New Zealand, Israel, France, Japan, South Korea, Brunei Darussalam, Italy, Spain, Kuwait, and the Bahamas. From an income perspective, Qatar holds the 6th rank among all 164 sample countries with more than USD 58,476 GDP per capita. Although its income level is comparable to the richest countries in Cluster 1, Qatar is classified in Cluster 2 due to its low environmental quality. The income levels of the other countries range within USD 55,653–29,375 GDP per capita for the 9th to 26th ranks, except for Spain, Kuwait, and the Bahamas, with USD 24,829, 23,929, and 22,830 GDP per capita in the 28th, 29th, and 31st ranks, respectively. Similarly, they show a matching pattern when it comes to the sectoral analysis. In the services sector, their value-added per capita ranges between USD 23,929 and 55,653 for the 8th to 28th ranks. Likewise, they have high value-added in the sector of industry, ranging between USD 18,906 and 4930, taking the 5th to 33rd ranks, except for Qatar in the 1st rank, and value-added per capita equivalent to USD 33,309. Correspondingly, in the manufacturing sector they show high value-added, mainly ranging within USD 2600–8250 for the 32nd to 4th ranks. In the agriculture sector, Iceland and New Zealand occupy the 1st and 2nd ranks, with USD 2739 and 1658 value-added per capita, respectively. The other countries also have high amounts of value-added, mainly ranging from USD 586 to 1064 from the 6th to 27th ranks. Therefore, Cluster 2 has high levels of income from both aggregate and sectoral-disaggregated perspectives.

Analogous to the economic aspect, the environmental scores are high in Cluster 2. From the EPI perspective, Denmark holds the 1st rank with a score of 82.5. After that, this score mainly ranges between the high values of 65.8 to 91.7 and high ranks of 5th to 28th in the majority for the other countries in this cluster. Regarding environmental health, Finland, Sweden, and Iceland have the 1st, 3rd, and 4th ranks with 99.3, 98.4, and 98.1 scores, respectively. The scores mainly range between 91.7 and 74.0 from the 8th to 28th ranks for the other countries, e.g., Denmark, the UK, Canada, France, The Netherlands, Japan, Germany, Austria, New Zealand, Spain, Belgium, and Italy. In terms of the ecosystem vitality score, Denmark occupies the 1st rank with a 76.4 score. This score ranges between 74.3 and 64.8 from the 4th to 18th ranks for other countries, such as the UK, Austria, France, Germany, Spain, Sweden, Finland, Japan, The Netherlands, and Belgium. Thus, both income level and environmental quality are high in countries assigned to Cluster 2, affirming that developed economies have high environmental quality, supporting the right side of the EKH and the PHH-Halo.

4.3. Cluster 3 (Other Countries)

Cluster 3 involves countries with low incomes and degraded environments, in contrast to those in Clusters 1 and 2. From the aggregate economic aspect, their GDPs per capita show the lowest values, mainly ranging within USD 375 and 27,089 from the 164th to 27th ranks, respectively. Regarding their sectoral evaluation, the services sector provides the lowest value-added to these countries, ranging from USD 155 to 14,285 for the 164th to 30th ranks, respectively. Similarly, the industry sector’s value-added mainly ranges within USD 27 and 4885 for the 165th to 34th ranks, respectively. This range is mostly within USD 9 and 2587 for the 164th to 33rd ranks in the manufacturing sector and within USD 39 and 963 for the 8th to 163rd ranks in the agriculture sector, respectively. Therefore, Cluster 3 includes countries with the lowest level of economic development.

Like the economic dimension, Cluster 3 indicates the lowest level of environmental quality. For the EPI score, these countries mostly indicate a range of 25.1 to 53.4, with the 164th to 45th ranks. For the environmental health score, this range mainly is between 11.8 and 73.0, with the 29th to 164th ranks. Similarly, the majority of these countries show a low range of scores between 23.9 and 53.8, with the 163rd to 45th ranks for ecosystem vitality scores, despite a few countries in this cluster having high scores. Hence, Cluster 3 involves countries with poor incomes and degraded environments, confirming the left side of the EKH and the PHH-Haven.

5. Discussion

This research categorizes countries into three clusters to reconcile three different hypotheses of the economy–environment nexus, the EKH, PHH-Halo, and PHH-Haven. Figure 4 illustrates the results by interconnecting the geographical location of the three clusters involving distinctive economic and environmental characteristics with the overlapping schema of the EKH, PHH-Haven, and PHH-Halo. According to the figure, countries in Clusters 1 and 2 with high level of economic conditions and environmental quality are located in North America, Europe, the Persian Gulf, the Far East, and the Pacific. The economy–environment nexus of these countries reflects the right side of the EKH. It also confirms the PHH-Halo. These results are consistent with [67,68] and confirm a positive connection between economic development and environmental development in the US and Europe, respectively. However, it is inconsistent with [69,70], which revealed a negative relationship between the economic and environmental factors in Singapore and Qatar, respectively.

The other countries, i.e., Cluster 3, which indicate low levels of economic and environmental status, are located in Asia, Africa, and South America. This cluster reflects the left side of the EKH. It also affirms the PHH-Haven. These findings support the results of [71,72,73], which indicate the negative relationship between economic and environmental factors in Asia, Africa, and South America, respectively. These results are persistent and robust, even with disaggregating the economy into different sectors (including services, industrial, manufacturing, and agriculture) and environmental quality into sub-indicators (including environmental health and ecosystem vitality). Therefore, the findings support the EKH, PHH-Haven, and PHH-Halo both at aggregate and disaggregate levels. These results affirm the nexus of the studied variables, aligning with [48,49], which accept “nexus” as a valid concept that effectively determines the behavior of the variables. However, this inference is inconsistent with [52], which proposes “nexus” as a buzzword unreliable in the real world.

The findings reveal a nonlinear connection between the economic and environmental factors from a correlational perspective, not within a causal analytical framework. They show that the economic and environmental variables follow a nonlinear correlation format, depending on the level of economic development, which is consistent with the EKH, PHH-Haven, and Halo. However, this revealed correlation does not show the direction of the relationship, as might be shown by conducting a causal analysis to infer if the economy influences the environment or vice versa.

Moreover, the results are reliable at an aggregate and global level, rather than a country- or industry-specific level. They categorize the majority of countries to show the whole and macro-behavior of economic and environmental elements at a global level. This behavioral and categorizing analysis demonstrates the common characteristics of countries at the most aggregated level, comprehensively covering all the environmental and economic indicators on an average measurement. However, it overlooks disparities in governance quality, regulatory frameworks, and institutional capacities at an intra-country level, obscuring nuances and masking heterogeneities within clusters, particularly in such a socio-environmental dataset with known structural inequalities. Thus, this global macro-analysis should not be generalized to a national level due to the fallacy of composition.

6. Conclusions

This research categorizes various countries in the world into three clusters, according to their economy–environment nexus. In this way, it tests the EKH, PHH-Halo, and PHH-Haven using machine learning and K-clustering techniques. The results indicate that countries with the highest economic development have the highest environmental quality. Similarly, countries with high economic development hold high environmental status. These results are consistent with the decreasing side of the EKH and the PHH-Halo. Furthermore, other countries with low levels of economic development show low environmental quality, reflecting the left half of the EKH and confirming the PHH-Haven. The results are similar even after disaggregating the economy into different sectors (including services, manufacturing, industrial, and agriculture) and the environment into sub-indicators (including environmental health and ecosystem vitality).

The findings have insightful implications for the literature by mapping the correlational typology of the environment–economy nexus at a global level, which must not be confused with a causal analytical inference. They confirm the nonlinear relationship between economic and environmental factors within an associative context, implying that the degree of economic development can determine how it correlates with environmental development. However, this inference should not be interpreted as the direction of this relationship; i.e., it is the economy that impacts the environment, or it is the environment that determines economic development. Therefore, the lack of causal relationship analysis can be regarded as a research limitation, underscoring a gap to guide future studies.

The findings include implications for international cooperation for sustainable global development among countries with different characteristics of economic and environmental conditions. They imply that developing economies should follow developed ones to remove structural barriers and reduce geopolitical asymmetries in terms of governmental, technological, and financial strategies. In this regard, developing countries should encourage foreign investment, particularly from developed countries, to attract advanced technologies with high efficiency and lower pollution. In addition, they should establish and strengthen their relationships with developed countries through regional and international treaties and agreements to enhance their green governance strategies. In other words, they should follow the patterns of developed economies to boost their economies while improving their environmental quality. They should focus on economic growth while strictly considering the environmental consequences. In addition, developed countries should maintain their current framework of simultaneously developing their economies and environment. Moreover, they should cooperate with developing countries not only to help them recover their economies but also to enhance their environmental quality.

This study concentrates on broad clusters, disregarding intra-cluster analysis. To fill this gap, future studies can examine the economy–environment nexus with higher resolution to explore the heterogeneities of countries in a cluster. Additionally, this research overlooks institutional elements such as good governance, environmental laws, and political stability, which can be the focus of studies in the future, as can investigating the association of economy and environment within a machine learning or clustering framework.