HMDA data is not just about collecting and reporting.  If used right, it can be a strategic asset that can produce bottom-line results.  I encourage you to make the most of your data by asking and answering questions based on HMDA results you see in LendingPatterns™. For example, you might want to further investigate penetration of low- and moderate-income census tracts in comparison to peers.

Lenders are familiar with the FFIEC methodology to determine statistical evidence of redlining.  The z-test is the standard statistical test used to determine if a lender’s market share difference between majority-minority tracts and majority white majority tracts is real.  When conducting redlining risk analysis, regulators look at the target lender as compared to its peers.  Specifically, they are evaluating against all lenders, lenders with ½ and 2 times the activity, and peers with similar business models for each major lending geography.

Lenders can use this same approach to assess low-income census tract penetration, for example.  This is a great way to show your regulator that you are providing qualified applicants with financing for homes.  Or conversely, this can give you warning that certain neighborhoods are being under-served.

To begin you will need a “Z-Score Calculator for 2 Population Proportions”.  A quick search on these terms in your browser yields numerous, free calculator results.  Upper income census tracts are the control proportion used for the analysis below.  So, you need the lending activity counts for low- and upper-income census tracts.  I found the mortgage applications information in table format in LendingPatterns™.  The premise is that, all things being equal, one would expect the subject lender’s market share (see the right-most column below) to be roughly the same across all tract income categories.

Z-Calculator Inputs and Results

 

 

 

 

Sample 1 Proportion total number of applications in low census tracts: 43

Sample 1 Size Market/Peer total: 6,760

Sample 2 Proportion total number of applications in Upper census tracts: 1,143

Sample 2 Size Market/Peer total: 96,840

Significance Level: 0.05

One would typically use the two-tailed hypothesis.

The value of the z-statistic is -4.0665. The p value is .00001. The result is significant at p < .05, and the market share is lower than the market share in upper income tracts.

This approach can also be used with consumer loan activity in these census tracts.  Subject lender and aggregate market data for small business and small farm loans can be easily accessed at the FFIEC.gov site under the Disclosure Report and Aggregate Report menu tabs.