APAN 5900 Solving Real World Problems with Analytics Predicting the Financial Risks of Real Estates Owned by REITs Affected by Hurricanes
Team Members: Raven Feng, Zixin Chen,Yicheng Shi, Zilin Zeng, Zehua Rong, Jingshu Yang, Aansh Mehta, Maddie Tremblay, Romauli Butarbutar
Background：
Standard & Poor Global Market Intelligence segment integrates research and analytical capabilities using data to answer clients’ questions. Through a comprehensive environmental performance profile, the Global Trucost segment analyses risks related to climate change, natural resource restrictions, and other ESG concerns. Driven by these two business interests, our project is mainly to help our clients to find the relationship between the financial impacts and climate change by collecting raw data and developing forecasting models for our clients. According to the prediction from the National Oceanic and Atmospheric Administration’s Office, when relative changes in hurricane activity are taken into account, average yearly losses are expected to rise by $7.3 billion, bringing the total cost of hurricanes and other coastal disasters to $35 billion per year. (NOAA) Real estate is particularly vulnerable to the effects of climate-related disasters since it is an illiquid, long-term investment. Therefore, it is important and constructive to begin keeping an eye on the effects of climate change on clients’ existing and target real estate assets in order to assess the risks. Despite the present availability of tools and resources to assist clients in assessing their exposure to these risks, many still struggle to factor the risk into their investment decisions. On one hand, it is difficult to quantify climate activities. On the other hand, people cannot rely on historical climate and weather data to predict future events, because we know that these patterns are changing. To help our clients to make more informed decisions when investing in the real estate sectors, our project aims to build a clear and actionable climate risk assessment framework that outlines the potential hurricanes risks and the real estate market. Before constructing the predictive model, we plan to use the historical data to find the relationship between climate activities and the real estate market.
Research Questions:
In this study, our major research question is “What is the relationship between hurricane intensity and the financial loss of impacted companies?” Below are three points that explain where was our data collected and how we built the data of hurricane and real estate assets. The first one is that hurricane intensity is measured by maximum wind speed per year and the number of hurricane encounters. The second one is that impacted companies are defined to be any companies within a 50-mile radius of hurricane landfall. The third one is that financial loss is defined to be asset write-downs.
Methodology:
We have four steps of the methodology for our analytical research. First, we explored and understood the structure of the data including what financial data is available over which years, how hurricane data is presented, features to join the financial data with hurricane data, and specific attention to geolocation (longitude and latitude of hurricane or property). For each storm, we have extracted its latitudes, longitudes, wind speeds, category, and year. We have then enriched the dataset in several ways. We’ve calculated the average latitude and longitude and extracted the maximum wind speed for each storm. Then, using the latitude and longitude coordinates, we’ve determined whether or not the storm fell within the US. We only want to consider storms within the US because we will be considering macroeconomic data that will be different across different countries - and we’ve specifically pulled the data for the US. Secondly, we have determined the scope of our research. We choose to use the existing data from 1995-2020. The data is divided into two sections:
- The data set of the real estate listed the public trade company.
- The data set of the hurricane in the United States.
Third, we have different ways to analyze each research problem. Our goal is to know the financial losses caused by the hurricane in the real estate industry. We will use macroeconomic data as control variables in our model, especially the GDP and the inflation rate -- to integrate external and non-hurricane market factors into our model. For each company, we pulled their institution name, the properties they own, their net book value per year, and the state where the property is located. This data can help us understand the change of the write-down value of the company's assets that have been impacted the most by hurricanes over time. We will establish a model of the key variables to see how the companies’ financials are changing in correlation with hurricane hits (and the strength of those hits). Finally, we will evaluate and improve the model iteratively based on model prediction ability and interpretability of the model.