Research Seminars at ILR
Here you find announcements of upcoming ILR research seminars given by external scientists across the full spectrum of the institute's research activities. Past seminars are also listed for information.
Regression coefficient estimation from remote sensing maps
Regressions are commonly used in environmental science and economics to identify causal or associative relationships between variables. In these settings, remote sensing-derived map products increasingly serve as sources of variables, enabling estimation of effects such as the impact of conservation zones on deforestation. However, the quality of map products varies, and -- because maps are outputs of complex machine learning algorithms that take in a variety of remotely sensed variables as inputs -- errors are difficult to characterize. Thus, population-level estimators from such maps may be biased. In this paper, we apply prediction-powered inference (PPI) to estimate regression coefficients relating a response variable and covariates to each other. PPI is a method that estimates parameters of interest by using a small amount of randomly sampled ground truth data to correct for bias in large-scale remote sensing map products. Applying PPI across multiple remote sensing use cases in regression coefficient estimation, we find that it results in estimates that are (1) more reliable than using the map product as if it were 100% accurate and (2) have lower uncertainty than using only the ground truth sample data and ignoring the map product. Empirically, we observe effective sample size increases of up to 17-fold using PPI compared to only using ground truth data. This is the first work to estimate remote sensing regression coefficients without assumptions on the structure of map product errors.
Jacob Moscona | (MIT)
Inappropriate Technology: Evidence from Global Agriculture
An influential explanation for global productivity differences is that frontier technologies are adapted to the high-income countries that develop them and "inappropriate'' elsewhere. We study this hypothesis in agriculture using data on novel plant varieties, patents, output, and the global range of crop pests and pathogens. Innovation focuses on the environmental conditions of technology leaders, and ecological mismatch with these markets reduces technology transfer and production. Combined with a model, our estimates imply that inappropriate technology explains 15-20% of cross-country agricultural productivity differences and re-shapes the potential consequences of innovation policy, the rise of new technology leaders, and environmental change.
Kathy Baylis |
(UC Santa Barbara)
Deep Learning for Predicting Counterfactual Deforestation in Protected Areas in the Amazon
The Amazon rainforest plays a critical role in absorbing billions of tons of carbon and is home to rich biodiversity. Yet, alarming rates of deforestation threaten its survival, with one-fifth already lost and more expected by 2030. Conservation efforts, such as establishing protected areas, aim to slow this loss, but assessing their true impact has been challenging due to differences in the locations chosen to be protected versus other forests, and the highly non-linear drivers of deforestation. In our research, we use a neural network trained on unprotected forest areas, to forecast deforestation. We then use these forecasts to predict what deforestation would have occurred without protection. We first predict whether the area would be deforested at all, and for those areas where we predict positive deforestation, we predict the amount of deforestation. Our findings provide new insights for evaluating conservation strategies, helping policymakers and environmentalists target interventions more effectively to preserve the Amazon and combat climate change.
Mengyu Liang | (Stanford)
Long-term commitment critical to reach forest restoration’s climate mitigation potential in East Africa
Forest restoration is gaining momentum as a natural climate solution to facilitate carbon dioxide removal while also addressing the biodiversity crisis. Globally, three primary restoration strategies—natural regeneration, assisted natural regeneration, and active restoration—have been adopted. However, inconsistent monitoring of forest dynamics means large uncertainties remain over their long-term potential to enhance carbon (C) removal. We examined over 30 years of forest aboveground carbon (AGC) estimates from high-resolution lidar, Landsat imagery, and field data across East Africa. Our study shows that assisted natural regeneration and active restoration outperform natural regeneration in enhancing forest C removal capacity, with long-term implementation (over 9 years) needed to overcome the initial lags in AGC accumulation. Restoring 14.24 million hectares of suitable areas available in East Africa, representing 2.1% of suitable restoration areas globally, could enhance forest C removal by 2.85 ± 0.82 gigatons of C (Gt C) by 2050. However, less than one quarter of this potential would be achieved by 2030.
Ashley Larsen | (UC Santa Barbara)
Spillover effects of organic agriculture on pesticide use on nearby fields
Organic agriculture is often suggested as a means to improve agricultural sustainability through less intensive and more natural production methods. Yet, the environmental impacts of organic product ion practices are only partially understood and whether such production practices have spillover impacts, beneficial or not, for surrounding producers remains unknown. Using a rich, field-level database of crop production and pesticide use from an intensive and diverse growing region in California, we seek to identify how surrounding organic cropland influences pest burden, proxied with field-level pesticide use, for both organic and conventional producers. We find surrounding organic agriculture leads to a small increase in pesticide use on conventional fields and a larger decrease on organic fields. This effect is effects primarily driven by insecticides, which is what we would anticipate based on prior ecological studies. Due to the different effects on conventional and organic focal fields, our simulation suggests that at low levels of organic agriculture in the landscape, levels commonly observed today, there is a net increase in pesticide use, which can be mitigated by spatial ly clustering organic cropland.
Céline Bonnet | (Toulouse School of Economics)
Can the economic and environmental impacts of policies targeting animal-based product consumption be reconciled? A value chain approach.
Given the significant environmental impacts associated with animal-based food consumption, this paper compares the effects of various policy instruments targeting consumers to reduce those environmental externalities. We evaluate price-based instruments-specifically, excise taxes linked to greenhouse gas (GHG) emissions and the removal of reduced value-added tax (VAT) rates-and contrast their environmental mitigation potential with policies that aim to change preferences for animal-based products.
In order to take into account the structure of the food supply chain and firms' price strategic decisions in the effectiveness of the policy instruments, we employ a structural econometric model of value chains. This model explicitly incorporates the market structures of food processing and retailing, as well as the interactions between these two stages. Using Kantar WorldPanel data on household-level purchases of branded animal-based products we first estimate consumption substitution patterns through a flexible random utility approach. We then estimate manufacturers and retailers' gross margins using a Nash Bargaining model of vertical relationships. We show that the disaggregation of animal products leads to higher elasticity estimates and that retailer margins are greater than manufacturer margins in most cases.
Those estimation results serve as a benchmark to simulate a range of policy scenarios -excise taxes, standard VAT rates, and preference-shifting interventions-targeting various categories of animal products, including beef, red meat, and all animal products. We show that tax instruments are overpassed on to consumers, thereby amplifying their environmental impact. Across all scenarios, food consumption's environmental footprint is reduced; however, these gains come at the cost of consumer welfare and profit reductions for both manufacturers and especially retailers. Nonetheless, in scenarios involving excise taxes or a VAT increase limited to red meat, the environmental benefits can outweigh the associated economic losses, suggesting that such targeted fiscal policies can be both effective and welfare-enhancing from a societal perspective.
Luke Sanford | Yale University
Satellite data and machine learning for causal inference
Understanding and properly estimating the impacts of environmental interventions is of critical importance as we work towards achieving global climate goals. Remote sensing has become an essential tool for evaluating when and where climate policies have positive impacts on factors like greenhouse gas emissions and carbon sequestration. However, when machine learning models trained to predict outcomes using remotely sensed data simply minimize a standard loss function, the predictions that they generate can produce biased estimates in downstream causal inference. If prediction error in the outcome variable is correlated with policy variables or important confounders, as is the case for many widely used remote sensing data sets, estimates of the causal impacts of policies can be biased. In this paper, we demonstrate how this bias can arise, and we propose the use of an adversarial debiasing model in order to correct the issue when using satellite data to generate machine learning predictions for use in causal inference.
Frederik Noack | UBC
Publishing in general interest journals
Discussion of differences between publishing in general interest journals, general economics journals, and various field journals, advantages and disadvantages, and some dos and don'ts.
David Schäfer | (Eurocare)
Challenges in implementing pesticide use in large-scale economic models for policy assessment
The reduction of pesticide use and their associated risks in agriculture is among the core goals of the Farm to Fork and Biodiversity strategy for 2030 and is also well embedded in the current reform of the CAP. A comprehensive economic and environmental policy impact assessment at European Union (EU) scale is still pending due to hurdles such as the representation of the large number of approved pesticides available to farmers. The aim of this seminar is to demonstrate current implementation developments of pesticide use and associated technology options for pesticide reduction in large-scale economic models with the example of the partial-equilibrium model CAPRI.
Jonas Schmitt | (ETH Zürich)
The role of on-farm crop diversification and its synergies with agricultural insurance to cope with extreme weather events in German agriculture.
Most existing weather risk assessments do not consider and compare the impact of multiple extreme weather events on yields of different crops at farm level. However, this is critical for farmers’ on-farm and off-farm risk management decisions and prioritization. Moreover, not all risk management strategies are effective or available for different severity level of damage events and farmers often need to combine different strategies to achieve an adequate risk reduction. For example, on-farm crop diversification may be sufficient for moderate, but not for extreme weather anomalies, and insurance options are often too expensive when designed to cover systemic risks such as drought. Therefore, this presentation aims to address the overarching question of how on-farm crop diversification and insurance can be combined to effectively exploit their synergies to improve cash crop risk management for different severity levels of weather anomalies.