cancel
Showing results for 
Search instead for 
Did you mean: 

Why the "middle score?"

Super Contributor

Why the "middle score?"

I've often read on here that when applying for a mortgage, the lender will pull your 3 common scores and use the middle one for their lending decision.  It's been about 15 years since I obtained my mortgage and I didn't even know what a FICO score was at the time, much less what my scores were.  Anyway, what is the reasoning behind using the "middle score" here?  Wouldn't an average be more indicative of where someone really stands?

 

For example, someone has scores of 700, 705 and 780.  Middle score is 705, average score is 728.  That difference could of course impact interest rate the loan is approved at.

 

Is it because in the above example, clearly there's a major difference between the high-score report and the other 2?  Perhaps the other 2 have a baddie present where the 780 score definitely doesn't... but since 2/3 of the reports contain the baddie, the lender considers those to be twice as "accurate" as the other, so the outlier (780) score gets thrown out, as does the 700, which of course doesn't really matter?

Message 1 of 19
18 REPLIES
Established Contributor

Re: Why the "middle score?"

We know that each report can have completely different information. Perhaps it's more about content than scoring. The middle likely isn't the dirtiest report nor the cleanest. But they see the content of all three anyway so I don't know.
Message 2 of 19
Frequent Contributor

Re: Why the "middle score?"

Unfortunately, no method is going to be perfect, really. While middle score may not be the most beneficial method (as opposed to using average or some other method) for some people, it might be the best method for others. Average could hurt some people just as easily as the middle score, it could be argued. Each method has its winners and losers.

 

Interesting fact (although you didn't bring it up): If you're going in on a mortgage with another person the mortgage lender will use the lowest middle score of the two of you.

FICO 8: EQ 784 / TU 786 / EX 791
FICO 9: EQ 797 / TU 797 / EX 793

BofA Cash Rewards Visa Sig $31,200 /Citi Double Cash $20,600 / CapOne QuickSilver $9,450 / Barclaycard Rewards $8,900 / Discover It $7,900 / Chase Freedom $1,500 / Lowe's $20,000
Message 3 of 19
Regular Contributor

Re: Why the "middle score?"

 

Why the "middle score?"

 

BrutalBodyShots:

 

     Actually there is quite an easy answer. You may remember from undergrad statistics that what you are calling the middle score is actually the median. The average (or mean) is the sum of the data divided by the total number of data points in the set as you said. The mean is only valuable whenever a graph falls on a normal distribution curve; however, when a graph (data) is skewed to one side, or there is a significant outlier, then the median is the only true measure. For example, if I want to find the middle income of a set of 5 people what would I use (mean or median)? What that means is I want to know what is a representative income level for a set of 5 people. Here are their incomes:

 

Person 1 makes $41,000 per year.

Person 2 makes $48,000 per year.

Person 3 makes $55,000 per year.

Person 4 makes $110,000 per year.

Person 5 makes $600,000 per year.

 

     The mean (average) is $170,800 a year. Ask if this average is really the “middle” salary of these five people. Of course not since 4 people earn less than this value and only one earns more. Now take the median ($55,000). Is this more representative? Sure it is because 2 people earn less and two people earn more than the median (middle) value in this example. So take your score example (700, 705 and 780). You got both the mean and median right (728 and 705); but look at the distribution. Is the mean (728) truly indicative of the middle for these values? Two values are BELOW the middle and only one is above the middle for the mean; BUT using the median, one value is below and one is above. If the data set has an even number the median is simple the mean of the two middle values (1, 1, 1, 2, 5, 6). Median is 1.50 (mean=2.67).

 

     Fundamentally, whenever there is a possibility of asymmetry in distribution or end values are not given, the median is helpful for measuring the location. Therefore, median is a better source of measuring central tendencies, especially if few values are clearly separated from the main body of the data (called outliers). That’s why the mortgage companies do it. Now on to quartilesSmiley Wink

 

Y

Current FICO 8 Scores:
EQ 792 (01/22/18), TU 799 (01/24/18), EX 824 (01/25/18)
EQ 794 (02/28/18), TU 801 (02/28/19), EX 820 (02/28/19)
EQ 794 (03/29/18), TU 799 (03/29/18), EX 812 (03/29/18)
EQ 789 (05/06/18), TU 797 (05/06/18), EX 809 (05/06/18)
EQ 797 (05/28/18), TU 806 (05/28/18), EX 814 (05/28/18)
Message 4 of 19
Super Contributor

Re: Why the "middle score?"

ytzak,

 

I appreciate your reply.  I think one of the issues though when discussing the mean/median topic would be total number of data points going into any given data set.  Using your example of the 5 salaries where 1 was a clear upper outlier, without additional data points is it really possible to tell if that 5th was really an outlier?  Maybe data points #6 and #7 (if they were obtained) are $350,000 and $425,000.  I agree completely with what you're saying if you're talking a solid amount of data.  5 salaries though IMO isn't a very big set and certainly 3 credit scores are an even smaller set to deal with, although with 3 major bureaus it's only possible to have 3 pieces of data to work with.

 

I agree with the previous post that there are likely benefits and drawbacks to all methods of taking an average.  Where the median may be a benefit for someone, it would be a drawback to another compared to taking the mean. 

Message 5 of 19
Regular Contributor

Re: Why the "middle score?"

     I completely understand your frustration, but let’s take my set and remove clear outliers. Let’s give the folks incomes of:

 

Person 1 makes $41,000 per year.

Person 2 makes $48,000 per year.

Person 3 makes $55,000 per year.

Person 4 makes $66,000 per year.

Person 5 makes $71,000 per year.

 

     In this case the mean is $56.2K, but the median is still $55K. Using the mean the curve still has 3 lower than the mean value and 2 above the mean. With the median it has two below and two above the middle. It is more accurate (on a bank’s statistical basis) and accuracy (to me) is always more fair.

 

     But here is something more important. This came up in another response I gave to another poster http://ficoforums.myfico.com/t5/Understanding-FICO-Scoring/Question-on-different-Experian-FICO-score... Why is there such an outlier in your example? There really should not be such a great difference in values (even with different FICO models) if all the data is the same in all CRs. In the post above, I had a situation where I had just such an outlier (on the low side). In my case (after a detailed review of my CRs) I discovered two errors that were affecting that score. In your example, maybe the same problem exists where there are errors on two of those reports and correcting it might pull both of the low scores up. If it is an error in your favor, well leave it alone until they catch and correct it themselves – then take the diveSmiley Wink.

 

Y

Current FICO 8 Scores:
EQ 792 (01/22/18), TU 799 (01/24/18), EX 824 (01/25/18)
EQ 794 (02/28/18), TU 801 (02/28/19), EX 820 (02/28/19)
EQ 794 (03/29/18), TU 799 (03/29/18), EX 812 (03/29/18)
EQ 789 (05/06/18), TU 797 (05/06/18), EX 809 (05/06/18)
EQ 797 (05/28/18), TU 806 (05/28/18), EX 814 (05/28/18)
Message 6 of 19
Valued Contributor

Re: Why the "middle score?"


@wrote:

I've often read on here that when applying for a mortgage, the lender will pull your 3 common scores and use the middle one for their lending decision.  It's been about 15 years since I obtained my mortgage and I didn't even know what a FICO score was at the time, much less what my scores were.  Anyway, what is the reasoning behind using the "middle score" here?  Wouldn't an average be more indicative of where someone really stands?

 

For example, someone has scores of 700, 705 and 780.  Middle score is 705, average score is 728.  That difference could of course impact interest rate the loan is approved at.

 

Is it because in the above example, clearly there's a major difference between the high-score report and the other 2?  Perhaps the other 2 have a baddie present where the 780 score definitely doesn't... but since 2/3 of the reports contain the baddie, the lender considers those to be twice as "accurate" as the other, so the outlier (780) score gets thrown out, as does the 700, which of course doesn't really matter?


Using a middle score is more likely to remove an outlier (either high or low) that could unduly influence score. 

 

It's kind of like income statistics. They typically report median, not mean as the highly compensated strongly influence the mean. When summary reports break income down into sub groups, it's often quintiles.

Fico 8: .......EQ 850 TU 850 EX 850
Fico 9: .......EQ 850 TU 850 EX 850
Fico 4 .....:. EQ 809 TU 823 EX 830 EX Fico 98: 842
VS 3.0:...... EQ 835 TU 835 EX 835
Fico 8 BC:. EQ 892 TU 900 EX 900
CBIS: ........EQ LN Auto 940 EQ LN Home 870 TU Auto 902 TU Home 950
Message 7 of 19
Super Contributor

Re: Why the "middle score?"


@     I completely understand your frustration, but let’s take my set and remove clear outliers. Let’s give the folks incomes of:

No frustration here at all, just trying to understand the reasoning behind the use of the middle mortgage score while ignoring the 2 scores on either side, that's all.

 

With 3 CRAs, it's possible that your middle score could be "wrong" [inaccurate] in either direction due to mis-reported information.  For example, one could have 1 dirty [legit] account.  That account may have only reported to 1 or 2 bureaus rather than all 3 the way it should be legitimately.  Scores may vary say 75 points because of this.  If it reported to just 1 bureau, the use of the middle score would be just fine.  If it reported to 2 bureaus, the applicant would be screwed.  This is a case where the median score could be significantly different depending on the reporting circumstance, where the mean would be more centrally located numerically.

 

I guess ultimately it doesn't matter, as all mortgage apps are going to involve a MR which will see all 3B reports, not just the "middle score" report.  I just find this topic to be interesting to think about.

Message 8 of 19
Regular Contributor

Re: Why the "middle score?"

     I completely agree BBS; but in all my statements I assumed the CRA data was correct. Technically, (as you said) if one CRA is not reporting a baddie, then their score is inaccurate and you should not benefit.  I agree they will do a merged report, but not always (especially with automated underwriting, with good scores and income). Technically, you should (in order to get a true score) report yourself. Yea right!

 

Y

Current FICO 8 Scores:
EQ 792 (01/22/18), TU 799 (01/24/18), EX 824 (01/25/18)
EQ 794 (02/28/18), TU 801 (02/28/19), EX 820 (02/28/19)
EQ 794 (03/29/18), TU 799 (03/29/18), EX 812 (03/29/18)
EQ 789 (05/06/18), TU 797 (05/06/18), EX 809 (05/06/18)
EQ 797 (05/28/18), TU 806 (05/28/18), EX 814 (05/28/18)
Message 9 of 19
Frequent Contributor

Re: Why the "middle score?"

I've found it to be pretty arbitrary, to use only the middle score, as well.

To wit; the underwriter is still going to require the applicant to pay off balances that may only appear on the report with the lowest score. This is a clear double standard.

It would make much more sense to just merge all of the accounts showing on all three major reports, and then produce a score from all of the data, excluding redundant accounts.

We certainly have the technological capability to accomplish this. However, I think it's all down to contractual limitations surrounding credit scoring.

Ultimately, it's still fair for everyone, as it stands, imperfect as it may be; because, by and large, everyone is still judged by the same standards.
Message 10 of 19