By Saul Fine is Founder, Chief Scientist & CEO of Innovative Assessments

More and more lenders around the world are using alternative data to improve their credit models and better serve the underbanked. Alternative credit data are typically derived from non-traditional sources of financial-related information, such as mobile/internet transactions and payments.

But these data can also be non-financial in nature, and as such, provide lenders with a very different kind of credit information. Non-financial data may include relevant aspects of a borrower’s personal character traits, for example, which are critical for good borrower behaviors. Traits such as responsibility and trustworthiness, for instance, are all needed to maintain good loan performance over time.

Indeed, such traits may be related to borrowers’ potential “willingness” to repay their loans, above and beyond their financial “ability” to repay. And this may be particularly important, given that one major reason for loan defaults is a refusal to pay. Similarly, we know from research in applied psychology that predictions of a person’s future behaviors should be based not only on objective situational factors, but on personal traits as well.


It certainly makes sense, therefore, that lenders should consider a borrower’s character in their credit decisions, even though traditional credit scores rarely address such factors directly. And this is not a new idea. In fact, today’s credit bureau scores were probably never intended to be one of the sole criteria for granting credit. Until relatively recently, for example, customers would typically apply for loans in-person, and loan officers would make subjective decisions based on many factors, including the person’s character. Moreover, “character” is one of the classic 5 C’s for assessing credit: Capital (savings), Collateral (securities), Capacity (income/debt), Conditions (terms of the loan) and Character (personal credibility).

Indeed, not considering character may cause traditional credit models to overestimate the risks of some of their low scorers, and underestimate the risks of some of their high scorers. In terms of the former, there are certainly many responsible and potentially valuable customers that banks and lenders are missing out on by relying primarily on financial-based scores. Not to mention the nearly 2 billion underbanked people globally who lack credit histories and/or prior use of banking services, upon which traditional credit models rely.

In terms of approved customers, character can help predict behaviors that credit histories cannot. Take, for example, a case of unexpected loss of income or increase in expenses. These life events are hard to predict, but how a person might behave in such situations – and the priorities according to which they might honor their loan commitments – are psychological in nature, and at least partly related to their characters.


Measuring credit-related character traits reliably can be tricky to say the least. Some solutions may tap various big data sources or social media as a means to infer insights about a person’s character, and this may indeed be informative. But such unstructured data sources can be difficult to model consistently across geographies and settings, and it may require very large datasets and long periods of time to train the data, and to empirically test their predicative values. Moreover, it can be hard to explain or defend the decisions made by such models.

Alternately, arguably one of the most effective ways to assess character traits is through psychometrics. Psychometrics is a science that has grown from more than a century of research and applications, but has only very recently been applied to credit scoring. Psychometric credit solutions take the form of self-report questionnaires, which ask key questions about a person’s typical financial behaviors, and use proprietary algorithms to profile patterns of responses to those questions. In addition, unlike big data, psychometric scoring solutions are based on underlying theoretical models whose measured constructs are explainable, and don’t require a lot of data training. Moreover, since psychometric tools don’t tap third party data sources, their systems do not require collecting personal identifiable information, and can thereby comply easily with the EU’s General Data Protection Regulation and similar data privacy requirements.

Finally, psychometric credit scores are not transient over time in the same way financial-based credit scores are. Personality and character traits are considered to be fairly stable over the years, and this can allow lenders to leverage a borrower’s psychometric score beyond the loan origination decision. Specifically, psychometrics can help lenders service existing accounts more personally, and even be proactive on collections, before a person goes into arrears. But to be sure, psychometric credit scores augment financial-based scores, and are designed to complement, not replace traditional models.


Psychometric solutions are not without their challenges, however. One challenge, for example, is that some borrowers may try to “game” the questionnaire by responding dishonestly. To mitigate this phenomenon, psychometric tools may include questions without right or wrong answers per se, as well as algorithms to identify insincere response patterns and response times, which can be used to disqualify such results.

Another challenge is that psychometric solutions require engaging the borrower, and this can add time to the loan application. On the other hand, considering that lenders may not need to administer questionnaires to applicants who might already pre-qualify, adding time to applicants who might otherwise not likely be approved is perhaps less of an issue. In addition, questionnaires need not be overly lengthy. Nowadays, some psychometric-based credit questionnaires can reach reliable scores in as little as three minutes.

One such psychometric solution is Worthy Credit by Innovative Assessments, where I serve as Chief Scientist & CEO. Worthy Credit asks borrowers to choose their preferences between equally desirable financial behaviors. For example, borrowers may be asked to endorse one of the following statements: “I organize my finances carefully” vs. “I avoid risky financial situations.” Responses to questions such as these indicate individual behavioral preferences, according to which psychometric score profiles are derived. Worthy Credit has scored over half a million borrowers since its launch in 2017, and has already been proven successful in 15 culturally diverse countries as a predictor of loan defaults, especially among the underbanked.

In all, psychometric-based credit scores show promise for the global lending industry, not only because they add a new type of alternative credit data not often considered, but also because they give lenders an opportunity to see beyond the purely financial data, and help identify more good borrowers.

Saul Fine is Founder, Chief Scientist & CEO of Innovative Assessments.

Image courtesy of GDJ.

This was originally posted on NextBillion’s website