Statistical Significance

This podcast explains how statistical significance is used in redlining allegations based on disparate impact, despite potential deemphasis under the Trump Administration, as regulators may shift accusations from disparate impact to disparate treatment while still relying on statistical analysis. The hosts clarify that statistical significance measures the probability that a bank's below-average performance in majority-minority census tracts occurred by chance rather than discriminatory practices, using a 5% significance threshold, and that larger banks with more loan volume must perform closer to market averages to avoid being flagged (ranging from 5% for 100 applications to 9.5% for 10,000 applications when the market average is 10%). However, the analysis emphasizes that statistically significant results can be misleading due to "lurking" or "confounding" variables, particularly when regulators use unrealistic market definitions (UREMAs) that include areas where banks lack branches or competitive presence, or when peer comparisons inappropriately mix different institution types like banks and mortgage companies—situations that have resulted in the majority of actual peer banks failing the statistical test, demonstrating the data was fundamentally skewed and making the statistical significance analysis unreliable. Brought to you by GeoDataVision and M&M Consulting