Michigan Department of Treasury
Predictive Modeling of Municipal Financial Instability
Background:
The Community Engagement and Finance Division (CEFD) monitors the finances of local governments. In recent years several local governments have experienced significant fiscal distress, and some have even required the assistance of an emergency manager. In order to better serve Michigan’s local governments, the CEFD is developing a system to identify communities at risk of fiscal distress before the situation becomes critical.
Project:
In order to build a model for predicting fiscal distress two datasets will be used. The first dataset will be the Annual Local Unit Fiscal Report (F65). This dataset has 1154 fields with records from the 2010-2017 fiscal years. The second dataset is the Audit Procedures Report (APR). This set has 42+ fields, of which some are yes or no questions. The APR records from are also from the 2010-2017 fiscal years. The CEFD has identified for four possible fiscal indicators. They are Unrestricted Fund Balance to Total Revenues Ratio, Cash Ratio, Net Asset Ratio, and Taxable Value Per Capita. The project objective is to train a model to predict fiscal distress based on past instances of local governments in fiscal distress using fields from the F65 and APR datasets.
Goals:
- Determine if past fiscal emergency could have been predicted prior to becoming critical.
- Determine if the four identified fiscal indicators are predictive.
- Identify new fiscal indicators.
