RiskExec has been updated to include the following enhancements:

Redlining Module

Redlining Analysis: Peer Detail Tab Updates

The count of branches for the target institution and the aggregate branch count of all lenders in a geography have been added to this tab. The branch counts include branches from the FDIC Summary of Deposits and from the Credit Union listing files.

The branch count in the Redlining Dataset row represents the count of branches for the Respondent ID listed in RiskExec for the Institution that user belongs to.

The number in the Geography Total row is the total count of branches of all institutions in the geography.

Fair Lending Categories

The RiskExec Fair Lending Categories have been added to the Redlining module. These fields can be used as filters, as well as fields in the Redlining Crosstab report under Dataset Reports.

Peer Analysis Module

Respondent Group Summary Report

Users can now run the CRA Respondent Group Summary report on the Average of Geography/Unit and/or the Total of Geography/Unit.

For all Respondent Group reports, for all data types, users can also choose more than four peers to run these reports on.

Fair Lending Module

Minority Borrower Distribution of HMDA Lending

This report has been updated to be grouped by Assessment Area. Previously, it was grouped by MSA.

Goodness-of-Fit Measures Added to Logistic Underwriting Regression Model Output

The following measures have been added to the Model Output tab:

• Model Convergence - Model convergence in logistic regression occurs when the algorithm adjusts the model’s settings until the changes become very small and stop improving, signaling the model is stable and ready to make predictions. A model is considered "converged" (TRUE) when it has found the best settings and no further improvements are made. If the model is "not converged" (FALSE), it means the algorithm hasn't found the best settings yet, and the changes are still significant, possibly due to issues like bad starting points, too many correlated variables, insufficient data, or incorrect settings in the model construction.
• Log Likelihood (NULL) - In logistic regression, this value is the Log Likelihood of the null model. It is used to calculate the Pearson Chi-Square.
• Pearson Chi-Square - In logistic regression, the Pearson Chi-Square test checks how well the model fits the data by comparing the observed outcomes to the expected ones. It is calculated using the Log Likelihood (NULL) and the number of Valid Observations. A low Chi-Square value means the model is a good fit, with the observed and expected values being similar. A high Chi-Square value suggests the model doesn’t fit well, as there's a big difference between the two.
• Hosmer-Lemeshow - The Hosmer-Lemeshow test checks how well a logistic regression model’s predicted outcomes match the actual results. It does this by grouping predictions and comparing them to what really happened. A low test value (closer to 0) means the predictions are close to the actual outcomes, which represents good model fit. A high value (closer to 100) means the model isn’t predicting well, which can indicate the model may not be a good fit. RiskExec uses 10 ordered queues to calculate this statistic. If the number of observations is such that 10 queues creates too few outcomes in each queue, then this statistic cannot be calculated and will display as NA.

Note: These new measures will only be available for analyses created after this release. Prior analyses will not have these measures unless they are re-run.

HMDA and Fair Lending Modules

New Field: Maximum Score

Maximum Score has been added as a calculated field in the HMDA module and in the Fair Lending module for the HMDA dataset type. If valid scores are present for both applicant and co-applicant in a HMDA record, the higher credit score between the two scores will populate in the new Maximum Score field. This field will automatically update if either of the two underlying scores are changed or updated.

HMDA, CRA, and 1071 SBL Modules

Distressed/Underserved Flag

The Distressed/Underserved flag has been added to the Property tab of the record detail page. 

This flag will update when geocoding is run on the file or record. 

The Distressed/Underserved flag has also been added to the File Crosstab report, as well as the Riskexec Full Data Export format, in all three modules. Users will also be able to filter their files using this field. 

Note: This flag will not populate until a file or record has been geocoded after this release. Until geocoding is run, the Distressed/Underserved flag will be set to X.