The Actuaries Climate IndexTM in Agriculture and Finance: Weather-Index Benchmarking and Unsupervised Weighting

Date

2026-7-1

Author

Yavrum, Cem

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

76
views

0
downloads

Cite This

Climate change poses significant challenges to the agricultural and financial sectors. This thesis comprises two complementary research studies that evaluate and extend the Actuaries Climate IndexTM (ACI) as a tool for measuring climate-related risk in actuarial valuation. The first study compares the explanatory power of the ACI with that of well-established weather-based indexes (WBIs) in crop yield prediction and weather-derivative pricing. The yields of corn, wheat, and soybeans are examined across six United States (U.S.) regions. Twenty-two model specifications are estimated using generalized statistical models and advanced machine learning algorithms. Principal component and functional principal component analysis address multicollinearity, and the results show that drought, wind, and sea-level components—alongside temperature and precipitation—materially affect yield variability. In the financial application, ACI components and WBIs produce broadly comparable derivative payoffs when they reflect similar climate conditions. Component-level models frequently outperform the equal-weighted composite ACI, indicating that equal aggregation may discard heterogeneous predictive information. Motivated by this finding, the second study extends the conventional equal-weighted ACI by introducing seven alternative unsupervised weighting approaches: EWM, SD, CRITIC, PCA, MI, COPULA, and MENT. The conventional and proposed indexes are evaluated comparatively using annual corn yield and quarterly agricultural GDP. Predictive performance is assessed under time-series cross-validation and ranked using a TOPSIS-based multi-criteria framework across six ACI regions and the aggregate U.S. The equal-weighted ACI remains a strong and transparent benchmark; however, on cross-regional averages, dependence-aware methods—notably COPULA and MENT for corn yield, and MENT and MI for agricultural GDP—consistently outperform it. Taken together, the thesis validates the standardized weather series aggregated in the ACI relative to traditional weather-based indexes and demonstrates that unsupervised reweighting of these series can improve composite-index informativeness without altering their definitions.

Subject Keywords

Actuaries climate index, weather-derivative pricing, crop yield prediction, weather-based indexes, machine learning, weighted ACI, unsupervised weighting

URI

https://hdl.handle.net/11511/120041

Collections

Graduate School of Applied Mathematics, Thesis