Suppose an entrepreneur or borrower with little credit or business history goes to a bank or some financial institution to borrow money.
Banks can use, as an alternative to a credit or business history, 'psychometric' or 'behavioural' tests.
What are some 'psychometric' or 'behavioural' tests used to gauge creditworthiness of a potential borrower with limited credit history? Where can I read more about these (eg some articles, what kind of articles or journals, keywords to google)?
Not really my area of expertise, but I find that a very interesting question and googled a bit for answers.
EFL (Entrepreneurial Finance Lab) is a for-profit company claiming to be the market leader in psychometric loan scoring. The company has been spun out of a research project of the Center of International Development at Harvard University. EFL's method has been covered in a number of media reports (e.g., here and here).
The basic idea is to apply psychometric data as the basis for determining whether to give micro-finance loans to would-be entrepreneurs in developing countries. In such emerging markets, traditional credit scoring used in industrialized nations are either not available or less reliable. If it works psychometrics might help to give loans to entrepreneurs who are less likely to default and thereby fight poverty in a more efficient way.
Given that this is a proprietary test, it is difficult to get detailed information about the kind of data that is used. According to the news reports and the company website, they assess fluid intelligence, certain attitudes and beliefs that are predictive of entrepreneurial success (e.g., locus of control), concrete business skills, and ethics and honesty. Part of these data rely on self-report questions, others seem use more behavioral data (e.g., performance in intelligence tasks or risk games).
To predict creditworthiness, they use Bayesian hierarchical modeling. From what I gather, initially, they used data from meta-analyses of predictors of entrepreneurial success as input (e.g., Rauch & Frese, 2007). Now, they seem to use their own data for model-building.
A source for more information on how this works may be a book that has been published by the founders of EFL (Klinger, Khwaja, & del Carpio, 2013). Citing from an abstract of chapter 2 of the book:
Regarding the Big 5 personality traits, extroversion is found to be strongly related to higher profit levels, with weaker relationships for agreeableness (positive) and conscientiousness (negative).
Interestingly, integrity is found to have a weak negative relationship with profits: the most honest entrepreneurs aren't the most honest (sic). Conversely, when considering default risk, the lowest risk entrepreneurs also tend to score higher on the integrity assessment, as well as register higher levels of conscientiousness.
Digit span (fluid intelligence), controlling for level of education, is negatively related to profit levels, but is not related to default risk.
When combined, these relationships with conscientiousness, honesty, and level of education have an AUC (a common metric of credit score predictive power) of 0.57 – 0.66, which is not extraordinarily strong when compared to credit scoring models in high - information countries and market segments, but it is sufficient to add significant value to the risk analysis task facing banks lending to SMEs in emerging markets.
We show that for one of the sample banks, risk of default for low-scoring clients is 50% higher than it is for high-scoring clients. Furthermore, we show that these results can be improved by customizing models to each country and financial institution, which isn't surprising given the cultural differences between Peru, Colombia, Kenya and South Africa. While traditional methods of model building suffer challenges of doing this customization without large amounts of data, new methodologies such as Bayesian methods are shown to offer promise to improve results even further, making customization without over-fitting possible and further strengthening the case for using psychometric tools for credit risk analysis.
An obvious question is whether this type of assessment can be gamed by malicious borrowers and loan officers (who tell their customers how they think they should answer to get the loan). EFL says they take various measures to prevent that (e.g., checking whether certain loan officers cause unlikely patterns of responses for their clients, randomization of responses, analyzing questions that have no obvious "right" answer, etc.).
The company claims to by highly successful in improving the quality of loan decisions by presenting a series of case studies. Nevertheless, given that this is a for-profit company and all analyses take place behind closed doors, it might be prudent to take this information with a grain of salt.
References
Klinger, B., Khwaja, A. I., & del Carpio, C. (2013). Enterprising Psychometrics and Poverty Reduction. SpringerBriefs in Psychology. New York, NY: Springer New York. Retrieved from http://link.springer.com/10.1007/978-1-4614-7227-8
Rauch, A., & Frese, M. (2007). Let’s put the person back into entrepreneurship research: A meta-analysis on the relationship between business owners’ personality traits, business creation, and success. European Journal of Work and Organizational Psychology, 16, 353–385. doi:10.1080/13594320701595438
