Effect of climate and geography on worldwide fine resolution economic activity

determinants: institutions versus geography and climate (some authors include also a third determinant, international trade or market integration [12, 13], which is however itself depen- dent on institutions, and geography). On the one hand, the popular book [14], and before it [4, 15, 16], have been arguing about the criticality of the role of institutions in driving productiv- ity. [4] attempted a quantification of the respective roles of institutions, geography/climate (the latter simulated by latitude only), but also integration, and concluded that once institutions are controlled for, integration has no direct effect on incomes, while geography has at best weak direct effects. However, because institutions are the result of a wider variety of factors, both exogenous and endogenous, including human, geography and climate factors, it is more difficult to disen- tangle causes and effects. For instance, as expressed in [17], one of the reasons why European settlers did not establish themselves in tropical areas (e.g. sub-Saharan Africa) is the presence of diseases such as yellow fever and malaria which are prevalent in tropical climatic conditions. On the other hand, it is difficult to identify appropriate institution-related indicators that are truly statistically independent of GDP. Of the six indicators considered by [18]–Voice and Accountability, Political Stability and Absence of Violence, Government Effectiveness, Regula- tory Quality, Rule of Law, and Control of Corruption–their individual linear correlation coeffi-

cient with GDP ranges between 0.6 and 0.9, and predominantly at the higher end of this range. For instance, Rule of law, as used in [4], has a coefficient of ca. 0.8 with GDP. More impor- tantly, however, a key limitation with institution indicators is that they are not available glob- ally at a resolution finer than country averages, which is essential for the work performed here. The most attractive feature of trying to explain economic development through geography and climate factors is that these are essentially exogenous elements. In fact, because weather is exogenous and random in most economic applications, it acts like a “natural experiment” thus allowing the identification of relationships between economic outcomes and meteorology in a scientific way [19, 20]. In this study we take advantage of a high-resolution global GCP data set [21]. The lack of a granular GCP global map has prevented researchers from intimately assessing the connection between geography/climate and economic activity. And although many attempts have been made [7 and references therein], studying their link at the level of country-average, i.e. using Gross Domestic Product (GDP), is sub-optimal. In fact, climate variables can vary wildly within a country. This is certainly true for very large countries like USA, China or Australia, but even for smaller countries like Italy, with a distinct north-south gradient in both climate and productivity, relying on GDP is not satisfactory [22]. Another important extension introduced here is the use of a non-linear model. Most of the

current literature simulates GCP variations using (multi)linear models, including [11]. Although practical and reasonable, linear models have major limitations, particularly in a com- plex non-linear problem like the modelling of economic activity. This study shows that climate and geography variables accurately explain, and reproduce, the per capita GCP worldwide at a fine resolution. In addition to latitude, less obvious variables such as month-to-month changes in air pressure are the next most important climate predic- tors. Our results provide a non-linear machine-learning framework for assessing and quantify- ing the main climate and geography factors affecting economic activity. These results, purely based on exogenous variables, have implications for understanding what could be favourable environments for thriving economies, or otherwise. The next section, 2, presents the data sets used, while the methodology is discussed in Sec- tion 3 and results are examined in Section 4. A summary and discussion are provided in Sec- tion 5. PLOS ONE Effect of geography on fine resolution economy PLOS ONE | https://doi.org/10.1371/journal.pone.0229243 March 2, 2020 3 / 20