Background

Credit scores are numerical scores, usally ranging from 0 to 1000, which reflect the credit worthiness of an individual or company.

Both credit scores and credit reports are designed so that a third party can meaningfully evaluate the strenght of an individual or company. There are some important differences though: credit reports are prepared by credit analysts who analyze financial data and assign letter grades which rank companies by their financial strength. On the other hand, credit scores are computed by statistical modelsthat use financial and business data to predict which companies are likely to become delinquent at some time in the future. In other words, credit reports are a credit analyst's best summary of a company's current condition, while credit scores are a statistical model's prediction about the future.

An important event in the emergence of credit scoring as an industry was the passage in 1970 by the US Congress of the Fair Credit Reporting Act or FCRA. Thirty years later, the three major credit bureaus (TransUnion, Experian, and Equifax) each maintain data on approximately 190 million Americans and use this data each day to create consumer credit scores. Consumer credit scores today are used for a wide variety of purposes, including direct marketing, creating instant automobile loans, and determining the points and interest rates of home mortgages.

Today, it is becoming more common to not only develop general purpose credit scores for consumers and businesses, but also to develop specialized credit scores for specific purposes (such as offering a particular type of loan or credit vehicle) and for specific vertical industries.

Characteristics of Credit Scores

There are three important characteristics of credit scores:

• Credit scores are determined by formulas or models. Credit scores are computed using mathematical formulas and statistical models. Broadly speaking, there are three internal components to the models. First there are inputs, which are derived from data about the individual or company. Second, there are parameters, which are used to weight the input data in various ways and to control the logic of the model. Third, there is a well defined way to combine the inputs and the parameters to create a score. This is sometimes called an algorithm.
• Credit scores are empirically derived. This means that credit scores are computed from data collected about individuals or companies. They do not reflect personal biases or information that is not quantitative. More precisely, this means that data is used to set the parameters of the model and as inputs in order to compute specific scores.
• Credit scores are statistically sound. This means that credit scores are derived using statistically valid procedures. More precisely, this means that the methods used to i) select an apppropirate statistical model; 2) estimate the parameters of the model; and 3) prepare the inputs of the model all are baed upon statistically valid procedures.

Computing Credit Scores

Credit scores are computed using a variety of different data, including:

• Trade or survey data. For individuals, trade data is collected about the timeliness of payments. For companies, trade data can also be collected or survey data can be used. For example, surveys can be used to capture the timeliness of payments from a company to its clients.
• Financial data. When available, financial data, and ratios derived from financial data, are also used to compute credit scores. Self reported financial data is treated differently than financial data reported by independent third parties or audited financial data. For small companies without audited financials, the credit score of the principals is sometimes used in liu of company financials.
• Other information. For an individual, the number of years that the person has lived in his or her current home, worked at his or her current employer, and similar information is generally used to compute a credit score. For a company, the number of years the company has been in business, the approximate number of clients the company has, and related information is also sometimes used.

Understanding Credit Scores

• Credit scores are designed to supplement and not replace credit reports prepared by credit analysts. Credit analysts can use a richer variety of data and put this data into context. Credit scores are not designed to do this, but rather to provide a simple, but effective, one dimension view of an individual or company.
• Credit scores are predictive. Credit scores assume that individuals or companies are divided into two groups: those that have had an adverse credit event and those that have not had an adverse credit event. Adverse events include going out of business, bankrupties, significant delays paying bills or clients, etc. The lower the credit score the more likely the individual or company is to experience an adverse credit event sometime in the future. It is important to note that credit scores are designed to predict what will happen in the future, not to summarize what has happened in the past.
• The statistical procedures used to collect data and to produce credit scores can be improved steadily over time. This is an important advantage of credit scores, and more generally, of statistically based business processes.
• Since many factors go into a credit score, since credit scores reflect how one individual or company compares to many others, and since the rules combining the factors can be quite complicated, explaining a particular credit score is not always easy. Credit scores are best understood as a statistical comparison of how one individual or company compares to other individuals or companies. In general, over time, those with higher credit scores tend to be more credit worthy than those with lower credit scores. Credit scores, like any predictions about the future, are not perfect. Sometimes individuals or companies with low credit scores do quite well (false negatives) and sometimes individuals or companies with high credit scores experience adverse credit events (false positives). Credit scores employ a number of quality assurance procedures to reduce the number of false negatives and false positives, but these are inevitable by the very nature of statistical prediction.

Three Reasons for the Growing Importance of Credit Scores

Here are three basic reasons credit scores are often preferable to credit reports.

Reason 1. Credit scores are based upon statistical models and not opinions. Credit scores computed with statistical models analyze historical data to make the best predictions about the future. This is in contrast to credit reports prepared by analysts in which the analyst's personal opinions may sometimes creep in.

Reason 2. Credit scores based upon statistical models are validated. With credit scores, data is carefully used to check the accuracy and performance of the predictions. This is called validation. In other words, you use validation to keep score on the model itself and you use this score to improve the model over time.

Reason 3. The statistical models used by credit scores can be improved over time. The underlying statistical model used by credit scores can be carefully improved over time as additional data is collected.