Podcast: The hidden math behind credit risk

Podcast

All Blog

The conversation covers why traditional machine learning models are missing critical components for accurate risk assessment and how adverse selection has dramatically impacted loan quality in recent vintages. Then Joe makes a bold prediction that software user interfaces are on the verge of a transformation that will render them unrecognizable from previous versions Curious? All is revealed in this fascinating conversation.

3 Views Read More

The evolution of goals in AI agents

All

Forced evolution has been proposed as a possible path to developing artificial general intelligence. For practical reasons, self-replicating robots are being proposed for missions where direct manufacture could be prohibitive or as a cost-effective means to maintain a stable working population of robots. If self-replication occurs in a harsh (i.e. selective) environment, the forces of evolution may distort the originally programmed objectives. Via millions of simulations of AI agents with nematode-level neural networks, this research explores the consequences of allowing replication in a hostile and competi- tive environment. As the selection pressures are tuned, the evolution of their neural networks and corresponding behav- ioral changes are tracked. As a consequence of these simulations, agents with multi-layer neural networks trained simply to retrieve resources, consume needed resources, and evade obstacles evolve behaviors that look like evasion of hostile overseers, the intended murder of enemies, and cannibalism of other agents. These simulations are intended to directly address safety concerns around creating self-replicating AI agents or robots. As designers, if we allow replication under selection pressure, regardless of initial designs, we risk allowing the emergence of unintended strategies. One solution to preventing evolution could be to enable AI agents with continuous backup– immortality.

16 Views Read More

An Age–Period–Cohort Framework for Profit and Profit Volatility Modeling

All

The greatest source of failure in portfolio analytics is not individual models that perform poorly, but rather an inability to integrate models quantitatively across management functions. The separable components of age–period–cohort models provide a framework for integrated credit risk modeling across an organization. Using a panel data structure, credit risk scores can be integrated with an APC framework using either logistic regression or machine learning. Such APC scores for default, payoff, and other key rates fit naturally into forward-looking cash flow estimates. Given an economic scenario, every applicant at the time of origination can be assigned profit and profit volatility estimates so that underwriting can truly be account-level. This process optimizes the most fallible part of underwriting, which is setting cutoff scores and assigning loan pricing and terms. This article provides a summary of applications of APC models across portfolio management roles, with a description of how to create the models to be directly integrated. As a consequence, cash flow calculations are available for each account, and cutoff scores can be set directly from portfolio financial targets.

17 Views Read More

Stabilizing machine learning models with Age-Period-Cohort inputs for scoring and stress testing

All

Machine learning models have been used extensively for credit scoring, but the architectures employed su er from a significant loss in accuracy out-of-sample and out-of-time. Further, the most common architectures do not e ectively integrate economic scenarios to enable stress testing, cash flow, or yield estimation. The present research demonstrates that providing lifecycle and environment functions from Age-Period-Cohort analysis can significantly improve out-of-sample and out-of-time performance as well as enabling the model’s use in both scoring and stress testing applications. This method is demonstrated for behavior scoring where account delinquency is one of the provided inputs, because behavior scoring has historically presented the most diculties for combining credit scoring and stress testing. Our method works well in both origination and behavior scoring. The results are also compared to multihorizon survival models, which share the same architectural design with Age-Period-Cohort inputs and coecients that vary with forecast horizon, but using a logistic regression estimation of the model. The analysis was performed on 30-year prime conforming US mortgage data. Nonlinear problems involving large amounts of alternate data are best at highlighting the advantages of machine learning. Data from Fannie Mae and Freddie Mac is not such a test case, but it serves the purpose of comparing these methods with and without Age-Period-Cohort inputs. In order to make a fair comparison, all models are given a panel structure where each account is observed monthly to determine default or non-default.

19 Views Read More

Classical and quantum computing methods for estimating loan-level risk distributions

All

Classical and quantum computing methods for estimating loan-level risk distributions

46 Views Read More

Effective Generative AI Model Risk Management

All

Effective Generative AI Model Risk Management

60 Views Read More

A Theory of Borrowers

All

Credit Risk, Adverse Selection, Age-Period-Cohort Models

63 Views Read More

Journal of the Operational Research Society

All

The new accounting standards of CECL for the US and IFRS 9 elsewhere require predictions of lifetime losses for loans. The use of roll rates, state transition and “vintage” models has been proposed and indeed are used by practitioners. The first two methods are relatively more accurate for predictions of up to one year, because they include lagged delinquency as a predictor, whereas “vintage” models are more accurate for predictions for longer peri- ods, but not short periods because they omit delinquency as a predictor variable. In this paper we propose the use of survival models that include lagged delinquency as a covariate and show, using a large sample of 30 year mortgages, that the proposed method is more accurate than any of the other three methods for both short-term and long-term predictions of the probability of delinquency. We experiment extensively to find the appropriate lagging structure for the delinquency term. The results provide a new method to make lifetime loss predictions, as required by CECL and IFRS 9 Stage 2.

12 Views Read More

Underwriting with Uncertainty: We're using ML wrong in underwriting.

Blog

There are a few more key topics I would like to share from my book... I believe one of the biggest errors in using Machine Learning models is not considering forecast uncertainty in underwriting.

126 Views Read More

Machine Learning: We're doing it wrong

Blog

I mentioned that this book started in response to a CRO's question. "We've done all the ML stuff and it hasn't moved the needle. What are we doing wrong?" I answered, "You're just building a fancier 1960s score. You have not integrated with cash flows, yield, and pricing."

129 Views Read More

Redesigning Credit Risk Modeling Chapter6

Blog

Chapter 6: The failure of risk-based pricing. This is where I get to mention that I've been observing how lenders set loan pricing for 30 years,

130 Views Read More

How IFRS 9 and CECL work better when using a panel data model

Blog

This week I was going to talk about how IFRS 9 and CECL work better when using a panel data model, but just days ago, Alan Forrest beat me to it with a discussion of how Simpson's Paradox arises when separate models are used for Stage 1 (12-month) and Stage 2 (lifetime) forecasts.

122 Views Read More

Redesigning Credit Risk Modeling Chapter4

Blog

In my book serialization, I promise the upcoming chapters get more interesting. Chapter 4 is the last part of the beginning. Ch 1: Introduction, Ch 2: What is the Goal, Ch. 3 Defining Yield, Ch. 4 Defining Volatility

119 Views Read More

Redesigning Credit Risk Modeling Chapter3

Blog

Continuing my serialization... Chapter 3: Profit Models

125 Views Read More

Redesigning Credit Risk Modeling Chapter2

Blog

I'm no Charles Dickens, but this is my attempt to serialize my book. Many chapters are just a few pages, so my summaries will be brief and give you the main point.

123 Views Read More

New Book: Redesigning Credit Risk Modeling

Blog

This is my little book of ideas about how to structure credit risk analytics to break free of the 1960s scoring paradigm and target the real problems in lending.

126 Views Read More

The evolution of goals in AI agents

Blog

Some research takes a long time to mature. The research for this paper involved years of algorithm development, running simulations, and developing new tracking metrics. Then there were the cycles of finding the venue for discussing this research.

127 Views Read More

Post Event with Allied Solutions LLC

Blog

I had the pleasure today of speaking about lending strategies for current uncertainties. This was a presentation hosted by our partner, Allied Solutions LLC, with a special focus on the credit union perspective, although the information is mostly general

154 Views Read More

Hello to my industry friends out there.

Blog

DFA has always worked well with partners to sell our products, but the diversity of what we have created goes beyond some of these niche relationships.

125 Views Read More

ChatGPT, Claude, and their many cousins

Blog

ChatGPT, Claude, and their many cousins are being much more widely adopted than you would guess even from the hype.

137 Views Read More

How do we monitor AI?

Blog

As we work with our first wave of clients for AI Monitor, www.deepfutureanalytics.ai, I am surprised at some of the use cases that are arising.

140 Views Read More