Scorecard Series: Visible Equity Database

New Call-to-action

Scorecard Series: Visible Equity Database

 For this four part series, Visible Equity would like to reintroduce a few cool new reports that cover the current credit union environment. We’re calling these new reports scorecards. We recently held a webinar where we unveiled these scorecards for the first time. You can watch the webinar and download the slides here.


In the webinar, we went over three scorecards. We wanted to create useful, complementary reports that give a simple, one page, snapshot summary of the current credit union space that are updated frequently. The first scorecard broke down loans and credit scores based on data found in Visible Equity’s database as well as a few key economic trends. The second scorecard reported trends from call report data. The third scorecard reported peer data also from call report data. The first three entries in this blog series will review these scorecards and go into more detail on the actual reported calculations. The fourth entry will go into a few trending peer statistics.


In this first entry, we’ll go over the Visible Equity Database Scorecard. Visible Equity collects tons of great data typically on a monthly basis, so we wanted to create something cool out of one of our greatest assets— our database.


Before jumping into the scorecard, we need to explain our sampling methods. Because Visible Equity does not contain every loan from every credit union in the country, we have to collect a sample that represents the population sufficiently. First, we separated loans by three types: autos, first mortgages, and credit cards. Within each of the three loan types, we used every credit union in the database that currently has those loans (not every credit union has all three loan types, but if a credit union did have a certain loan type, it was included in that loan type’s sample). Next, we decided to sample the number of loans from a credit union based on each credit union’s state population. Each state received a weight that was defined as the state’s population divided by the US population. We then divided each state population weight by the number of credit unions that belong to each state within the database and that possess a given loan type. For example, California has the highest population of all states, so it’s weight is the largest. However, for each loan type, our database contains many credit unions from California, so when we divide the state weight by the number of credit unions from California in the database, the weight given to each credit union in California is smaller. If a state was not represented, we distributed that state’s weight evenly amongst all other states’ weights.


With that in mind, here is our database scorecard.



The majority of the data for this scorecard come from February 2017 data.  The top half is separated by three loan types: autos, first mortgages, and credit cards. The bottom half covers new and used autos, applications, and economic trends. We’ll explore each section.


Above we have the top section. Along the left, we can see how the statistics are broken down by loan type. When creating the samples for this report, we obtained sample sizes of 200,000 auto loans, 75,000 first mortgages, and 200,000 credit cards. To the right of that column, we have the average and median credit scores by loan type. Since the average is less than the median, we know that the distribution of credit scores is left skewed, which means there are some outliers toward the lower end. From these we see that first mortgages typically have the highest credit scores, followed by credit cards, and then autos.


Next, we see the median LTVs for autos and first mortgages. Autos come in high at 80% while first mortgages come in at 52%. The next column is the median loan age. First, the age of the loan is calculated as the number of months between the “current” date of February 2017 and the loan’s origination month, and then it is converted to years. We then found the median of all loans within the sample for each loan type. Credit cards currently typically have the longest lives, followed by first mortgages, and then autos.


The next three columns are shaded in a pale orange because these metrics come from NCUA call report data. For consistency, we used the call report data from the credit unions used for the samples. A big difference between these columns and the Visible Equity database columns is that these data are only updated quarterly. The call report data for this scorecard come from Q4 of 2016. The Visible Equity data can and will be updated more frequently (monthly).


In the first orange column, we have the median interest rate where autos come in lowest (and are distinguished between new and used), followed by first mortgages, and then credit cards. The next column is the Days Past Due percentage. This is calculated by dividing the amount of loans that are 60 or more days delinquent by the total amount of loans by loan type (when we say amount, this is in terms of dollars). All three loan types show low DPD% hovering around 0.5% with credit cards being the highest, followed by autos, and then first mortgages. The last column, CO%, is the median annualized charge-off percentage, and this is calculated by dividing the amount of year-to-date charge-offs by the average of the current quarter’s total amount of loans and the previous year end’s total amount of loans by loan type and multiplying this percentage by 12 divided by the last month of the current quarter (so for Q4, you would multiply the percentage by 12/12=1).  All three loan types are very low, with autos being the highest, followed by credit cards, and then first mortgages, which is expected given our knowledge of delinquencies. Note that even though the first mortgage median CO% is 0%, this does not mean there were no charge-offs during Q4. This just implies that at least half of the credit unions in the sample did not experience first mortgage charge-offs (there were, in fact, first mortgage charge-offs in Q4).


The next section of the report covers credit score grade distributions and migration. For each loan type, we found each loan’s credit score and assigned each loan a letter grade based on the following scale.








  >= 720  

  720 > & >= 680  

  680 > & >= 640  

  640 > & >= 600  

  600 > & >= 550  

  < 550  


For each loan type we see bar graphs displaying the grade distributions below. Clearly for all three types, the vast majority of loans have A+ credit scores, and the percentage decreases for each decrease in letter grade. One takeaway is that there is a great disparity between A+s and all other grades for first mortgages. But this makes sense. Since first mortgages are typically larger loans, institutions will require borrowers to have stronger, positive borrowing behavior to extend the loans.


Below each bar graph is the credit score migration where “=” stands for “no change,” “+” stands for “improved,” and “-” stands for “declined.” For each loan, we took the original credit score grade and determined whether the grade changed or not and in which direction. For example, if a loan originated with an A and currently has an A+, that loan is considered to have improved. Clearly, most grades have not changed for all three loan types, but a sizable portion have improved and declined with credit cards (showing the most improvement) and autos (showing the greatest percentage of declining grades).


The next section breaks down autos and applications. In the autos section, we have new autos on the left, which make up about 25% of auto loans, and we have used autos on the right, which make up about 75% of auto loans. In the bar graphs, we show the percentage of each auto type that are direct loans (in blue) and indirect loans (in orange). For new autos, slightly more than half were direct loans, while slightly less than half were indirect loans. Used autos, on the other hand, show closer to two-thirds direct versus one-third indirect.


Below autos we have a breakdown of applications data. For each loan type we have the percent of applications that were approved, the percent denied, the look-to-book ratio, and the approve-to-book ratio. The look-to-book ratio is the number of applications that were funded divided by the total number of applications. The approve-to-book ratio is the number of loans that were funded divided by the number of applications that were approved. It’s important for us to point out that the sample size used for this applications section was significantly smaller than all other samples used in this scorecard, but we expect to gather more data soon to improve our sample.


In the applications bar graph, we see that most applications are approved (in blue) for autos and credit cards, which means that a small proportion of applications are denied (in orange) for those loan types. However, notice that first mortgages show a slightly higher denial percentage relative to the approval percentage. The look-to-book ratio is lowest for first mortgages, then autos, and then credit cards (in gray). And finally, for the approve-to-book ratio, we see the same order of first mortgages coming in the lowest, followed by autos, and the credit cards (in yellow).


Finally, we have a plot of economic trends that do not come from the Visible Equity database. These are public data made available by the Federal Reserve Bank of St. Louis and Zillow. In blue we have Zillow’s price-to-rent ratio for the US. This ratio is found by taking the estimated median price of a home in the US and dividing it by the estimated median rent for a home in the US. From the plot, we see that the peak of this trend occurred right before the recent financial crisis, which dramatically decreased in the following years, and we are currently experiencing an increase in the price-to-rent ratio again. This plot is a cool indicator of how the housing market is performing. If we were to go back in time to 2006 and see how high the plot had reached, we might think twice about purchasing a home.


The orange line is the US unemployment rate. Note that this is in terms of a percentage. As expected, the peak occurred around 2009 and 2010, and it looks like we are currently back to low levels, which shows an improvement in the economy.


Finally, we have the 30-year fixed mortgage rate average (also in terms of a percentage), which has generally declined over the past 20 years. We are currently seeing a little uptick in that rate, though.


So there we have it. This first scorecard gives a nice little overview of the current lending state of credit unions. We plan to update this scorecard about two weeks after the end of each month. In the next installment of this blog series, we will be looking at call report trends.

Keaton Baughan

Product Manager