Working Papers
Nguyen, Q., Diaz-Rainey, I., Kitto, A., McNeil, B., Pittman, N., & Zhang, R. (2022). Scope 3 Emissions: Data Quality and Machine Learning Prediction Accuracy, SSRN Working Paper
This paper explores the quality of Scope 3 emission data in terms of divergence and composition and the performance of machine learning models in predicting Scope 3 emissions. We do so using the Scope 3 emission datasets of three of the largest data providers (Refinitiv Eikon, and ISS). We find considerable divergence between third party providers, making it difficult for investors to know their ‘ exposure to Scope 3 emissions. Surprisingly, divergence exists between the datasets for emissions values that have been reported by firms (identical data points between Bloomberg and Refinitiv Eikon). The divergence is even larger for ISS when it adjusts reported values using its proprietary models ( identical data points). With respect to the composition of Scope 3 emissions, firms generally report incomplete compositions, yet they are reporting more categories over time. There is a persistent contrast between relevance and completeness in the composition of Scope 3 emissions across sectors, as irrelevant categories such as travel emissions are reported more frequently than relevant ones, such as the use of products and processing of sold products We also find that the application of machine learning algorithms can improve the prediction accuracy of the aggregated Scope 3 emissions (up to 6%) and its components, especially when each category is estimated individually and aggregated into the total Scope 3 emissions values (up to 25%). It is easier to predict upstream emissions than downstream e missions. Prediction performance is primarily limited by low observations in particular categories, and predictor importance varies by category. We conclude that users of the Scope 3 emission datasets should consider data source, quality and prediction errors when using data from third party providers in their risk analyses.
Presentations [* co-author]:
- International Association for Energy Economics Webinar (Online, 2022)
- The 4th JRC Summer School on Sustainable Finance (European Commission - Joint Research Centre & Online, 2022)
- The 5th Annual GRASFI Conference (University of Zurich, Switzerland, 2022) GRASFI Best Paper Award for Climate Finance Research
- Banco de España Webinar (Online, 2022)
- Sustainable Finance and Investment Seminar (Stanford University & Online, 2022)*
Nguyen Q., Diaz-Rainey, I., Moore A. B, Thorsnes. P., Cox, S., McKenzie, L. & Stik-Wang L. (2019) ‘Risk to Residential Property Values from Climate Change-Related Flooding Hazards: A Mixed Method Approach’, SSRN Working Paper
Greater South Dunedin (GSD) has been identified as one of the most vulnerable areas to Climate Change-Related Flooding Hazards (CCRFH) in New Zealand, yet little is known about the magnitude of how CCRFH will impact property values. We address this issue by proposing a novel modelling strategy that links CCRFH, and in particular Sea Level Rise (SLR), to of residential property value, at fine geographical resolution. The strategy is both empirical and forward looking modelling. The empirical analysis reveals a significant negative price effect for houses associated with flooding risks in the local market (between 5.9% and 3.1%), which existed prior to the June 2015 South Dunedin flood and was exacerbated temporarily after this event. The forward modelling projections apply a “bathtub fill” approach to elevation data in a Geographical Information System (GIS) to identify the properties that will be inundated (using IPCC scenarios to 2100). Uncertainties arising from data error and long-term projection are modelled through Monte Carlo simulation. We find that, the risks of permanent inundation are currently limited and only become non-negligible in a “business-as-usual” pathway. The risks of periodic flooding, however, are strikingly large across all scenarios in the presence of extreme events. With high tides, the number of inundated properties may be as high as 39%. With extreme rainfalls, this number potentially increases to 41%. Taken together, CCFRH may affect property worth up to NZ$ 983 million in rateable value (37% of GSD property market). We conclude by acknowledging the limitations of our “bathtub fill” approach.
Presentations [* co-author]:
- GRASFI 2019 Conference (University of Oxford, 2019) - shortlisted for the Union Bank of Switzerland (UBS) Innovative Methods Prize
- GIS @ Otago Symposium (University of Otago, 2019)
- The Planning Team (Dunedin City Council, 2019)
- The 12th OERC Energy & Climate Change Symposium (University of Otago, 2018)
Bodeker, G., Cox, S., Diaz-Rainey, I., Moore A. B, Nguyen, Q., Thorsnes. P., (2022). Sources of Uncertainty in Asset-Level Modelling of Change-Related Flooding Risk to Residential Property Values. Work-in-Progress.
Analyses of climate change-related flooding hazards and the related impacts on individual residential property values require accurate projections of flooding risks into the future. Such projections should be nuanced enough so that the extent and timing of potential flood hazards can be mapped to exposure and vulnerability at the property level. These estimations are subjected to uncertainty arising from the lack of knowledge into the futures – represented by the range of climate scenarios that reflect the possible evolutions of climate. The modelling process involves the combination of several data layers (for instance, projection of sea-level rise, rising groundwater level, extreme sea level), the data uncertainty inherent to these layers may affect flood hazard projection accuracy (Cooper & Chen, 2013; Cooper et al., 2015; Sweet et al., 2017). Further, several architecture choices are to be made during the modelling process which may well have a substantial impact on the estimation results. The choices include, but are not limited to: the choice of land elevation models, the thresholds or proxies for hazard (the mean sea level or the mean high water spring), the choice of inundation models that either assumes a horizontal groundwater level everywhere (the ‘bathtub’ fill model) or factors in subsurface heterogeneity and spatial variations in water table height (the groundwater model), the identification of the “zero baselines”, the spatial resolutions on which inundation maps are generated, the damage functions to be employed, as well as the number of Monte Carlo realisations to be ran. It is important to understand how climate scenarios, architecture choices and data uncertainty may affect the projections of future flood hazards and the flow-on effect on residential property values. In this paper, we highlight the importance of uncertainty estimates when they relate to high impact public-interest decision-making. We do so in the context of modelling the change in coastal and low-lying flood risks due to sea-level rise of the Greater South Dunedin area, one of the most vulnerable areas subject to climate change in New Zealand (PCE, 2015).
Presentation:
- The 4th New Zealand Geospatial Research Conference (Massey University, 2022)*
- Center for Disaster Resilience Seminar (University of Twente, 2022)
- The Green and Sustainable Finance & Institut Louis Bachelier Seminar (Paris, 2022)
- Climate Change, Insurance, Finance & Housing Workshop (University of Auckland, 2022)*
- Assembly of Investment Chairs (Auckland, 2022)*