Suraj R. Nair

Global climate change is widely expected to reshape patterns of migration, but the nature of those effects are not well understood. We use data from the world's largest social media platform --- covering roughly 3 billion individuals --- to estimate the impact of long-run changes in climate on migration at multiple scales. The impacts are nonlinear and unevenly distributed, increasing migration along certain corridors but generally constraining mobility in the poorest parts of the world. Using modern climate models to project these impacts into the future, we estimate that climate change will internally displace millions of individuals by the end of the 21st century, while also inducing widespread immobility in large parts of the world. We find no evidence to suggest that warming will increase migration to wealthy nations; if anything, these results suggest a net decline

Seasonal migration plays a critical role in stabilizing rural economies and sustaining the livelihoods of agricultural households. Violence and civil conflict have long been thought to disrupt these labor flows, but this hypothesis has historically been hard to test given the lack of reliable data on migration in conflict zones. Focusing on Afghanistan in the 8-year period prior to the Taliban’s takeover, we first demonstrate how satellite imagery can be used to infer the timing of the opium harvest, which employs a large number of seasonal workers in relatively well-paid jobs. We then use a dataset of nationwide mobile phone records to characterize the migration response to this harvest, and examine whether and how violence and civil conflict disrupt this migration. We find that, on average, districts with high levels of poppy cultivation receive significantly more seasonal migrants than districts with no poppy cultivation. These labor flows are surprisingly resilient to idiosyncratic violent events, including extreme violence resulting in large numbers of fatalities. However, seasonal migration is affected by longer-term patterns of conflict, such as the extent of Taliban control in origin and destination locations.

Digital loans have exploded in popularity across low- and middle-income countries, providing short-term, high interest credit via mobile phones. This paper reports the results of a randomized evaluation of a digital loan product in Nigeria. Being randomly approved for a loan (among those who otherwise would have been denied) substantially increases subjective well-being after three months, but being randomly approved for a larger loan does not have any effect. Neither intervention significantly impacts other measures of welfare, and we can rule out large impacts -- either positive or negative -- on income and expenditures, resilience, and women’s economic empowerment.

Over 30 million people globally consume illicit opiates. In recent decades, Afghanistan has accounted for 70–90% of the world’s illicit supply of opium. This production provides livelihoods to millions of Afghans, while also funneling hundreds of millions of dollars to insurgent groups every year, exacerbating corruption and insecurity, and impeding development. Remote sensing and field surveys are currently used in official estimates of total poppy cultivation area. These aggregate estimates are not suited to study the local socioeconomic conditions surrounding cultivation. Few avenues exist to generate comprehensive, fine-grained data under poor security conditions, without the use of costly surveys or data collection efforts. Here, we develop and test a new unsupervised approach to mapping cultivation using only freely available satellite imagery. For districts accounting for over 90% of total cultivation, our aggregate estimates track official statistics closely (correlation coefficient of 0.76 to 0.81). We combine these predictions with other grid-level data sources, finding that areas with poppy cultivation have poorer outcomes such as infant mortality and education, compared to areas with exclusively other agriculture. Surprisingly, poppy-growing areas have better healthcare accessibility. We discuss these findings, the limitations of mapping opium poppy cultivation, and associated ethical concerns.

As the major ingredient of concrete and asphalt, sand is vital to economic growth, and will play a key role in aiding the transition to a low carbon society. However, excessive and unregulated sand mining in the Global South has high socio-economic and environmental costs, and amplifies the effects of climate change. Sand mines are characterized by informality and high temporal variability, and data on the location and extent of these mines tends to be sparse. We propose to build custom sand-mine detection tools by fine-tuning foundation models for earth observation, which leverage self supervised learning - a cost-effective and powerful approach in sparse data regimes. Our preliminary results show that these methods outperform fully supervised approaches, with the best performing model achieving an average precision score of 0.57 for this challenging task. These tools allow for real-time monitoring of sand mining activity and can enable more effective policy and regulation, to inform sustainable development.

We have developed flexible machine learning algorithms which can detect construction-grade sand and gravel resources in river basins and coastlines at global scale with high spatial resolution (10 m). Our approach uses object based image analysis methods fusing freely available Sentinel-1 and Sentinel-2 multispectral satellite datasets. This method achieves an F1 score of 87.5\% and accuracy of 88.71\% using a random forest classifier trained on a newly aggregated global dataset of in-situ grain size observations. We further validate performance in sections of the River Ganga where a gravel to sand transition is known to occur, and in a section of the River Sone where a number of known sand mining concessions exist. This work lays the foundation to build end-to-end deep learning models that can predict where illegal sand mining occurs.

At the core of many consumer lending decisions is a credit score: an algorithmic assessment of a customer's creditworthiness. Traditional credit scores are designed to maximize lender profits, and use machine learning algorithms to predict which customers will repay loans. This paper proposes and tests a different paradigm for consumer lending, in which `welfare-sensitive' credit scores allow the lender to balance expected profits against the expected welfare impacts of specific loans. Using data from a randomized control trial in Nigeria, we show how machine learning algorithms can be trained to predict the welfare impact of lending to a client, and how those welfare scores can be combined with traditional credit scores to characterize a Pareto-efficient tradeoff between welfare and profits. Our main result suggests that, in the Nigerian context, the lender could achieve an 11\% gain in consumer welfare by sacrificing 0.1\% of profits.