Navigate to next or previous posts:

« Thoughts about Democracy in America | Emergent Chaos Main page | The Purple States »

You talk like a delinquent

(Posted by cwalsh)

This is interesting. Not sure how robust the finding is, but according to an analysis of LendingClub data on all past loans, including descriptions of the use for the money, applicants using certain words in their descriptions are much more likely to default.

For our purposes define a Delinquency as either being late in your payments or having defaulted completely. The 10 words with the greatest p-values are below. [...]


WordLoans WithP(Delinquency|No word)P(Delinquency|Word)p-value
also
215
0.067
0.140
0.0004
need
608
0.062
0.105
0.0015
business
233
0.069
0.116
0.0038
live
91
0.070
0.154
0.0057
already
64
0.071
0.156
0.0059
other
285
0.068
0.112
0.0081
bills
223
0.067
0.135
0.0082
bill
279
0.066
0.125
0.0117
interest
660
0.081
0.053
0.0136
"Words and Credit Scores", Social Science Statistics Blog

Not something I've studied, but I wonder if a neural network could successfully classify these loans?

Posted by cwalsh on November 3, 2008 at 11:48 AM in Economics , Science . You can: comment, view comments (3), search Technorati.

Bookmark this post:

Comments

You should be able to use an off-the-shelf Baysian spam classifier for this. That said, there are certain spam words with much higher P-values than any of these!

Posted by: Nicko | November 3, 2008 1:16 PM

"I need the loan to refinance my mortgage and buy Viagra"

:^)

Posted by: chris | November 3, 2008 1:53 PM

You have obviously been using my patent p-value enlargement products :-)

Posted by: Nicko | November 5, 2008 5:54 AM