CreditForecast.com provides insight into how the U.S. and local economies affect consumer credit behavior and performance. This unique tool provides history and forecasts for a wide range of household credit, economic, and demographic variables at a detailed level of geography enabling you to examine, segment, and stratify credit risk and economic data across states, metropolitan areas, and rest of states.
CreditForecast.com provides historical credit data, updated monthly, back to July 2005. Forecasts have a five-year horizon and a quarterly periodicity, and are updated on a quarterly basis.
CreditForecast.com is a joint product of Equifax and Moody's Analytics, leaders in the collection, analysis, and forecasting of household credit trends. Discover What's New »
CreditForecast.com subscribers will be able to participate in quarterly conference calls devoted to a discussion of current and anticipated trends in household credit conditions. Each call begins with a brief description of the current economic environment and baseline forecast and includes an assessment of risks to the baseline forecast. We then discuss recent and forecasted trends in consumer spending, balance sheets, and credit usage. Each call lasts 60 minutes and includes an accompanying PowerPoint presentation.
For more information on signing up for a Consumer Credit Outlook Briefing, contact your Equifax consultant or Moody's Analytics at 1-866-275-3266. Outside of the U.S. or Canada, please call 610-235-5299.
You asked and we listened. In response to customer feedback, Equifax and Moody's Analytics have added loan origination vintage and risk score dimensions, as well as monthly updates of credit data.
at 1-866-275-3266 to ensure you are getting the maximum benefit from these amazing features.
CreditForecast.com can be used by lenders to benchmark their portfolio's performance against that of the entire credit market.
This can be done across product lines and detailed levels of geography and for both the growth in credit and credit risk. Historically, lender benchmarking has been done solely vis-à-vis specific competitors. CreditForecast.com allows for benchmarking against the entire market, which will allow lenders to identify risks and opportunities that cannot be determined by benchmarking against a self-identified set of peers.
CreditForecast.com can be used by lenders to improve the predictive ability of their current customer-level scoring.
Lenders have historically used credit scoring models based only on customers' past credit history. Since economic conditions change quickly and are often different than those that prevailed historically, credit decisions based on credit scores developed using historical information do not fully account for all the information available to lenders. CreditForecast.com provides lenders with a method to incorporate forward-looking economic information into credit decisions.
CreditForecast.com can be used to improve the accuracy of P&L forecasting.
Most P&L forecasting done by lenders at a portfolio level does not take into account the changing economic environment. These models are often based on historical roll rates that remain constant regardless of economic conditions. Roll rates are highly dependent, however, on the health of the job market and broader economy. P&L models in which roll rates are a function of economic and credit variables available from CreditForecast.com thus result in more accurate forecasting.
CreditForecast.com can be used to evaluate the sensitivity of lenders' portfolios to shifting economic conditions.
Just how important are falling unemployment or rising interest rates to loan growth and credit quality? Do they vary significantly across product lines and regions of the country? Lenders can determine whether, for example, a strongly expanding economy experiencing lower unemployment, but higher inflation and interest rates will have a negative impact on the credit quality of their mortgage versus credit card portfolios.
CreditForecast.com can be used by regulators to assess the performance of financial institutions.
Certain lenders may be at heightened risk of suffering credit losses, given their regional and product line exposure. This risk can be measured using CreditForecast.com. Mortgage lenders in California, for example, may be at greater risk in a rising rate environment given surging activity and housing values in recent years. Lenders can also use CreditForecast.com to demonstrate to regulators their ability to measure and manage risk.
CreditForecast.com can be used by investors evaluating opportunities and risks in the rapidly growing asset- and mortgage-backed securities markets.
Do interest rate spreads on ABS and MBS pools adequately reflect current and prospective economic and credit conditions? Are interest rate spreads appropriately discounting regional economic and credit differences across pools? Equity investors can also use CreditForecast.com to evaluate earnings growth prospects and stock price valuation of lending institutions.
CreditForecast.com can be used to determine the size of potential markets a lender is interested in entering. Users can identify areas for expansion which fit their target customers’ economic and credit profiles. This can be achieved by comparing trends and forecasts in existing locations to potential markets. Information broken down to the MSA and rest-of-state level provides users with great flexibility in analyzing different geographies.
To ensure detailed coverage of all major types of consumer credit, CreditForecast.com provides credit data for seven mutually exclusive product categories and 20 subcategories. Users can license all seven product categories for the full picture of U.S. consumer credit, or they can license just the categories that are most important to their business.
Economic forecasts are based on the work of over 60 economists, nearly two-thirds
of whom hold an advanced degree, who monitor and forecast the U.S. economy at the
global, macro, regional and industry level.
Forecasting is done for origination risk score bands, geographic areas, and vintages,
with the total forecast being additive. The forecast database includes a baseline
most probable forecast and four alternative scenarios. The alternative scenarios
vary according to where we are in the business cycle. The forecasts are updated
with new history and forecasts quarterly. The economic and demographic forecasts
are produced using Moody’s Analytics’s detailed econometric model
system. View full details on all Moody's Analytics Macroeconomic Outlook
Alternative Scenarios (must be logged in to view).
The credit forecasts are produced with a model developed using a pooled time series,
cross sectional specification. Each equation is estimated for all metro areas using
the available historic data. Projections of the credit, economic and demographic
variables are available with a monthly periodicity and a five-year forecast horizon.
Forecasting Methodology and Alternative Scenarios
The forecasting models for CreditForecast.com consist of structural vintage-level
models of aggregate consumer credit behavior. Specifically, the models predict
loan balances, performance, utilization and average credit scores across each of the lines of business. Models are developed on aggregated cohorts of loans broken down
by lines of business, vintage and geography. CreditForecast.com generates time-series
forecasts for all performance series across all lines of business within 250
geographies and four consumer origination credit score groups over the next five
Explanatory variables for forecasting loan performance may be defined by three broad
categories: vintage age or maturation, economic environment at time of origination,
ongoing or time varying economic environment. Forecasting accuracy is the main criteria
used to specify and accept models though attention is also paid to econometric stability
and theoretical consistency of models.
Vintage Modeling Overview
The general approach to modeling the performance of individual loan cohorts (defined
as the overlap of geography, vintage, original credit score and product category)
considers three main drivers, Lifecycle, Vintage Quality, and Prevailing Conditions.
These underlying drives impact each of the key components of the way retail credit
* Cohort: overlap of geography, vintage, score and product category.
Forecasting New Vintages
One of the advantages of the new version of CreditForecast.com is that it is vintage
based. Not only does this allow users to compare historical performance across origination time periods,
but future vintages are forecasted as well in order to generate forecasts for any of the time series.
For example, consider the forecast of the 60 DPD rate in March 2013. % 60 DPD is
defined as $ Balance of loans 60 DPD in March 2013 / $ outstanding balance at the
start of March 2013. While the numerator may come mainly from loans already in existence
as of March 2012, the denominator will be made up of loans that already exist as
well as new originations through March 2013. Therefore both performance of the existing
book and future originations need to forecasted in order to develop a complete picture.
Forecasting Changes in Volume
Economic factors were chosen to capture the default and prepayment incentives facing
borrowers in aggregate. Balances should decrease quickly when rates are low
and borrowers can refinance. Similarly, balances may be hypothesized to decrease
when house prices are falling (a proxy for general economic health) and more loans
are subject to default. Seasonality plays a role in the time of year of the
reporting period — more mortgages are paid off in the summer, for example,
as borrowers move. Time of year of origination also appears to have a permanent
effect on loan performance as loans that are originated off-cycle tend to perform
worse than those that are originated on-cycle.
Having estimated the period to period changes in volumes, a time series can then
be computed for outstanding originations by computing balances for the runoff portfolio
as well as for future loan originations. Change in balance equations are estimated
both in terms of dollars and number of accounts.
Credit Performance Measures
All of the performance measures for delinquency and termination status are converted
to conditional percentages by dividing by either the outstanding balance or number
of accounts in each time period. For example, the 30 day delinquency rate
is computed as:
30 DPD Rate (Number) = # 30 DPD accounts / # active accounts
30 DPD Rate (Balance) = $ 30 DPD accounts balance / $ active accounts balance
Vintage level economic variables are measured at the start of each vintage period and remain
constant through the life of each vintage.
Other Credit Measures
A number of summary measures beyond balances and performance
are included in the CreditForecast.com service, including aggregate payment and
high credit amounts. These measures are converted to a per account basis by dividing
by the number of active accounts in each geography-vintage-origination credit score
cell and then modeling these per account values.
As the historical time series show
little variation on a per account basis, models for payment, high credit and per
account balance are estimated as a function of geography, origination credit score
and economic factors.
Performance rates are modeled as these are stationary and directly tied to
economic variables. For example, the unemployment rate is highly correlated
with delinquency rates but less correlated with the actual number of delinquent loans, as
the latter fail to account for the size of the outstanding pool or denominator.
Number of trades and dollar balances are computed for each performance category
by multiplying the forecasted performance rates from each model by the forecasted
balance and number of accounts from the relevant model. Utilization rates
are computed by dividing the high credit dollar balance by the dollar balance outstanding
in each cohort.
Historical DataThe historical credit data are made available from the Equifax National Consumer Database, which contains information collected from over 12,000 contributors who furnish data on tradelines, collections, public records and demographics across a broad range of industries, with updates on a daily basis.
Within a month, the database experiences on the order of 1.7 billion tradeline updates, 120 million collection updates, 350,000 public record updates, 185,000 bankruptcy updates and 12 million address changes. Information being added to the database goes through a rigorous quality assurance process, including automated and manual reviews, to ensure the highest accuracy possible.
The credit data are sourced from the full Equifax credit database and include consumers with >=1 active trade. This results in approximately 226 million consumer records in a given month. Selection always occurs at the end of the month, so that the results are not affected by any in-month reporting fluctuations. Monthly historical files starting with June 2005 were used. Therefore, full “open-to-close” vintage data are available starting with July 2005, while information on trades in older vintages – when available from the consumer credit files – is also included.
Equifax developed special credit trend attributes that aggregate the consumer file into attributes specifically intended for monthly trending. These attributes consider only trades with activity in the last three months and only look at current delinquency statuses in order to focus on the true state of credit (balances, high credit, delinquencies, etc.) in a given month. These attributes were calculated for each consumer file and then aggregated to the ZIP code, county, MSA and state levels. This will allow CreditForecast.com to quickly adapt to any future changes in county or MSA definitions.
Origination and current risk score level classification are done at the consumer level, based on ERS 3.0 – Odds Scale (score range 280-850). There are four score bands, allowing the user flexibility in identifying consumer risk profiles of interest. Forecasting is done separately for each of the score bands; the total forecast is additive.
Comparison With Other Data SourcesThe historical CreditForecast.com data have a number of significant advantages over publicly available sources of credit data. These other sources include data on consumer loan delinquencies from the American Bankers Association, mortgage loan delinquency and foreclosures available from the Mortgage Bankers Association, and consumer and mortgage delinquency and charge-off data available from the Federal Reserve Board and FDIC.
The most significant advantage of the CreditForecast.com data is that they are available for the nation, states, and metropolitan areas. The ABA and MBA data is available only for the nation and states, and the Federal Reserve Board and FDIC data is only available for the nation.
The CreditForecast.com data are also timelier than the other sources of credit data. They are made available within four weeks of the end of the month, compared to two to three months after the end of the quarter for the other sources.
The CreditForecast.com data is also based on where the borrower resides. The ABA data, by contrast, are based on the branch or operating location of the lending institutions participating in the survey. This may or may not coincide with where the borrower lives. This is of particular importance for the ABA’s data on bankcards, given that many of the cards are originated by lenders with operations in New York, Delaware and South Dakota to borrowers that live all over the country. The MBA does ask lenders to report by the state where the property is located.
Since the CreditForecast.com data are based on borrower information, they are not affected by changes affecting lenders. The ABA and MBA data can be significantly affected by any change in the lenders included in their surveys. For example, the MBA has previously expanded the number of subprime mortgage loans in its survey, resulting in a substantial increase in measured credit problems. The Federal Reserve and FDIC data, which are based in part on call reports provided by lending institutions, can also be affected by changes in the status of lenders. Historically, the acquisition by a large commercial bank of a retailers’ credit card portfolio and the resolution of failing lending institutions have affected measured credit quality.
Forecast Data Projections of the credit, economic and demographic variables are available with a monthly periodicity and a five-year forecast horizon.
Forecasting is done for the origination and current risk score bands separately; the total forecast is additive. The forecast database includes a baseline, most probable, forecast and four alternative scenarios. The alternative scenarios vary according to where we are in the business cycle. The forecasts are updated with new history and forecasts four times per year.
The economic and demographic forecasts are produced using the detailed econometric model system of Moody’s Analytics. A description of this system is available upon request. The credit forecasts are produced with a model developed using a recursive pooled time series, cross-sectional specification. Each equation is estimated for all metro areas using the available historical data. The equations are log-linear in form, meaning the terms (unless otherwise indicated) are specified as the natural logarithm and the coefficients can be interpreted as elasticities. All the equations contain metro area-specific constant terms to capture the metro area-specific fixed effects. Finally, all of the equations contain an autoregressive term to capture the inertia in the series.
Consumer Credit Data SeriesFor each of the categories, historical data and forecasts are available at the national, Census region, Census division, state, and Metropolitan Statistical Area (MSA) levels for the following variables:
Historical and forecast data are available for several consumer credit risk and bankruptcy scores. Mean Credit Scores for the nation, regions, divisions, states, and MSAs are provided.
Blended VariablesCreditForecast.com includes unique blended series created by merging economic data with credit data. These blended series are available for historic data and forecasts, and for all geographies.
The Loan to Value Ratio (LTV) is the measurement of aggregate mortgage balances outstanding to the value of the housing stock. Homeowners’ Equity per Household is the difference between the home value and mortgage debt for an average household. These series provide information on homeowners’ real estate wealth, the equity cushion against a housing downturn, and the size of real estate debt.
The Equity Extraction to Income Ratio measures the extraction of equity on existing homes as a percentage of disposable income. This series is calculated using a similar methodology developed by former Chairman Alan Greenspan and Fed economist James Kennedy. Equity extraction is the amount homeowners are withdrawing from their housing wealth, potentially to be used for home improvements, consumption and debt restructuring.
The Debt Service Burden is the minimum amount households must pay to service their debt as a share of their disposable income. The regional series is constructed using a methodology similar to that used by the Federal Reserve in constructing national estimates and is benchmarked to be consistent with that series nationally. The debt service ratio is an accurate measure for households’ financial burden, although it is sensitive to changes in homeownership rates.
Definitions are as follows:
As detailed in the model documentation, prior to 2015Q1 the econometric models embedded in the CreditForecast service
relied on ordinary least squares regressions in modeling performance variables with appropriate variable transformations,
controls and restrictions to insure rates fell within the [0,1] range. Over time, econometricians have developed alternative
techniques for estimating rate models on cohort-level data which have more robust statistical properties such as the fractional
logit (see for example Papke and Wooldridge (2008) ). This approach is better suited both theoretically and empirically to handle
proportional data and can handle zero-values in a more elegant and robust fashion than the linear regression model. As a result,
the model specifications were switched over to the fractional logit starting with the 2015Q1 release. While the impact on top-line
forecasts should be minimal, users may observe some improvement in the forecasts for more granular cuts of the data which are more subject
Under the implicit assumptions of the fractional logit model, reported rates of zero and non-zero are assumed to evolve from the same data
generating process with independent and identically distributed (iid) errors following the binomial distribution. Alternative, robust
methodologies are used in reporting standard errors to account for these assumptions.
It is important to note that foreclosure is often a complicated process involving
borrowers, loan servicers, lawyers, courts, recording offices, sheriffs, mortgage
insurers, and other interested parties. As a result, the paperwork trail is long,
and the status of a loan at any given time may be ambiguous depending on the source
of the information. For example, a servicer may classify a loan as being "in foreclosure"
if a letter has been sent to the borrower indicating that legal proceedings will
be initiated as a result of the borrower's failure to pay his obligation. However,
the court system will not indicate a foreclosure until formal legal briefs have
Data on mortgage volume and performance available from CreditForecast.com are compiled
directly from Equifax's database of consumer credit reports. With virtually all
mortgage servicers reporting information on the performance of their loans to the
major credit bureaus, this data set provides the most comprehensive and timely information
on the current status of all mortgages in the country directly from servicers. Although
Equifax has specific reporting guidelines that servicers are to follow when reporting
the status of their loans, with more than 12,000 individual loan servicers in the
U.S., there is the possibility for discrepancies in the data resulting from inconsistent
treatments across servicers, as well as lags in reporting.
According to Equifax guidelines, servicers are to report the current delinquency
status of each loan and the foreclosure status in separate fields (for example,
"Bob Johnson's loan is currently 180 days delinquent and in foreclosure"). Some
servicers may only report the delinquency status without the foreclosure field.
Further complicating matters is the possibility for a borrower to transition out
of foreclosure status back to paying status (for example, a borrower may work out
a loan modification with the servicer before foreclosure proceedings advance to
the point of no return).
As a result, the foreclosure data presented on CreditForecast.com have great utility
in examining trends by origination vintage, credit score band, and state/metropolitan
area, but the reported levels may not match with government sources. The combination
of loans that are more than 120 days delinquent and those that are in the foreclosure
process provides a high-quality measure of all loans that are distressed and likely
to complete the foreclosure process within the next 12 months.
To arrive at the attributes, Equifax aggregates account information, as defined
above, for qualifying accounts for each consumer file that passes the filtering
criteria, for each of the attributes below.
Referred to as "severe derogatory" in the prior version of CreditForecast.com, the
term default is used interchangeably with the terms loss and write-off. Default
includes charge-off, foreclosure, repossession, and defaulted student loans. Default
is a gross concept; recoveries are not netted out. Bankrupt accounts are not included.
Default accounts reflect a negative final disposition (other than bankruptcy) in
the current period. This is in contrast to the closed positive status, which reflects
that the account became closed in the current period and is not delinquent. Bankruptcy
is another final account disposition.
Accounts that are open and at least 120 days past due, in collections, or have started
the foreclosure process are included in the 120+ DPD or collections category.
Because the defaults are flow measures, a default rate in a particular month can
be annualized directly by multiplying the defaults (units or volume) in a particular
month by 12 and then dividing by the units or volume in that month.
To calculate a 12-month default rate for a particular cohort of loans, take the
sum of defaults (units or volume) for 12 consecutive months and divide by the sum
of all active accounts (units or volume) at the beginning of the 12-month period.
Active accounts are defined as current through 120+ DPD.
High credit is defined as the loan amount/current credit limit on the account if
it is reported by the lending institution; otherwise, it is the highest balance
that was ever reported on the account. For most installment loans, the series is
the original loan amount (principal only). For revolving accounts, the series is
the current credit limit on the account. However, some lending institutions do not
provide credit limit information for competitive reasons, in which case it is the
highest balance ever reported.
As such, the measures of credit utilization (balance divided by credit limit) and
credit limits in CreditForecast.com are a mix of true credit limits and highest
balance (high credit). This is standard industry practice. Scores, which run off
of these data, use the high credit for calculation of credit utilization in the
absence of a credit limit, as this information cannot be separated.
Student loans in a deferment period should continue to be reported each month by
the lender. The loans will be reported as deferred and carry a status of current.
The CreditForecast.com attributes do not exclude deferred accounts as inactive.
The "inactive" criteria are used to remove accounts that have not been reported
in the last three months in order to exclude information, such as balances or statuses,
that is not up to date, as opposed to removing accounts that are not actively in
use. For example, a credit card that has not been used will still be included as
long as the lender is reporting it. Such an account would show a zero balance and
a status of current.
No. Equifax systematically selects a single consumer to represent the account in
the case of jointly held accounts. This means all of the consumer-level attributes
tied to the account (such as risk level, geography, balance, etc.) will be represented
by one consumer.
The following information, as well as other information reported by the contributing
institution, is contained on the credit file for each account. Access to fields
depends on the individual/institution viewing the file and the file format chosen.
Equifax uses a combination of the lending institution's reporting industry code
(for example, Automotive Dealers-New, Credit Unions, Mortgage Cos.), account type
(installment, revolving, open), narrative codes (e.g., Credit Card, Home Loan, Line
of Credit), as well as other account information to classify each account. Additional
filters are applied to the data to minimize classification errors. As a result,
misclassification of loans should be negligible.
To maintain consistent origination vintages in CreditForecast.com, a trade's product
category is set to how that trade is classified at its first entry into the CreditForecast.com
dataset and is then maintained for the life of that trade. Therefore, any future
reporting changes from the reporting institution will not alter the original product
category, as it could in the prior version of CreditForecast.com.
Manufactured housing loans that are reported to Equifax by data contributors are
included in the data. The majority appear in the First Mortgage product category,
but some get classified in the Home Equity Loan product category, depending on how
the loans are reported to Equifax. A classification enhancement was made for CreditForecast.com
(4.0), so that all Agency installment mortgage products are classified as First
Mortgage â€“ Agency.
A consumer is defined as an individual with a credit file. Deceased consumers are
even included in the credit data, in order to keep the origination vintages whole.
When summarizing counts of consumers, the following rules are applied:
Accounts are assigned to risk score bands according to the risk score of the borrower
in the month prior to the open date of the account (origination) and in the current
Accounts are assigned to cohorts of accounts that were opened during the same time
period, or vintages. By adding this extremely valuable dimension to the data, the
unique performance of vintages can be tracked and compared over time. A number of
steps are taken to ensure the integrity of the vintages.
Equifax and Moody’s Analytics strive to continually improve the CreditForecast.com
solution, whether prompted by customer requests or internal learnings and developments.
The breakout of the Auto Loans and Leases data was a result of both of these factors.
Customers from the Auto market space expressed a need for the separation of Auto
Loan and Lease data within CreditForecast.com, and Equifax developed a method to
accurately identify Auto Loans and Leases in the Equifax credit database.
The breakout reveals insights into the unique volume, credit composition and performance
characteristics of Auto Loans vs. Auto Leases over time, and aligns the data with
how lenders typically conduct analysis of their auto finance portfolios. The
more granular view facilitates more precise and pertinent benchmarking exercises,
and it’s especially useful as a data supplement input into portfolio performance
forecasting and stress testing, like that done within Moody’s Analytics’
Credit Cycle service.
While there are no specific guidelines for the reporting of balance on an Auto Lease
trade, Equifax research indicates that many Auto Lease data contributors report
the Balance as the product of (Monthly Payment Amount x Remaining Number of Months
in the Lease Term).
Over 99% of the accounts and 97% of the balances in the Retail Product Category are revolving
accounts with close to 100% unsecured. The average balance for this product category is $318
while the High Credit is $2,038. The majority of accounts in the Retail Product Category originate
from Clothing Stores, Department Stores, Mail Order Firms, Jewelers and Oil Companies.
82% of the accounts and 46% of the balances in the Consumer Finance Product Category are revolving accounts
while 18% of the accounts and 54% of the balances are Installment accounts. 92% of the accounts in this Product
Category are unsecured. The average balance for the Consumer Finance Product Category is $1,180 while the High
Credit average is $3,905. The majority of accounts in the Consumer Finance Product Category originate from Sales
Finance and Personal Loan Companies.
The Other Product Category consists of accounts that could not be classified in any other Product Category with the
information provided on the account (i.e. First Mortgage, Home Equity, Bankcard, Auto etc.) 52% of the accounts and 19%
of the balances in the Other Product Category are revolving accounts while 48% of the accounts and 81% of the balances are
installment accounts. The average balance for the Other Product Category is $7,506 while the High Credit is $12,001.
Over 93% of the accounts in this Product Category originate from Banks, Credit Unions and Savings and Loan Companies.
In response to customer feedback, a number of significant and valuable enhancements
have been made to the new CreditForecast.com. These changes collectively result
in more granular, relevant and actionable data, which promote better-informed business
decisions and strategies. The enhancements are as follows:
As a result of the significant enhancements made to the service, particularly the
improved sampling techniques, the new CreditForecast.com credit data is not "apples-to-apples"
comparable to the CreditForecast.com Classic credit data. There are additional finer-detail
changes that may cause accounts to be included or excluded differently across the
two versions of CreditForecast.com. Current CreditForecast.com Classic users are
encouraged to assess the impact of the changes on their business applications and
determine if any adjustments are necessary.