Who Is Most Likely to Default on a Loan?

DATA STORY

Who Is Most Likely to Default on a Loan?

What 21,454 borrower records reveal about household profile, income, and loan purpose

A descriptive credit-risk analysis prepared for online publication

July 2026

Default risk is rarely explained by one characteristic. A review of 21,454 cleaned borrower records shows meaningful differences across household profile, income band and loan purpose—but also demonstrates why percentages need context before they become lending rules.

The headline findings

·        The overall portfolio default rate is approximately 8.1%.

·        Unmarried borrowers record the highest marital-status rate at 9.8%.

·        Car loans have the highest purpose-based rate at 9.4%, closely followed by education loans at 9.2%.

·        High-income borrowers have the lowest income-band rate at 7.1%.

·        The number-of-children pattern is uneven and does not rise steadily with every additional child.

Children and default: the pattern is not a straight line

Borrowers without children show an observed default rate of 7.5%. The rate rises to 9.2% for one child and 9.5% for two children, then drops to 8.2% for three children. The four-child group reaches 9.8%, while the five-child group records no defaults.

Figure 1. Observed default rate by number of children.

The small groups require caution. Only 41 customers have four children and just 9 have five. A percentage based on so few records can shift sharply with only one additional default. The evidence therefore does not support a blanket claim that every additional child steadily increases risk.

Marital status shows the widest large-group spread

Unmarried borrowers have the highest observed rate at 9.8%, followed by customers in civil partnerships at 9.3%. Married borrowers record 7.5%, divorced borrowers 7.1%, and widow or widower customers 6.6%.

Figure 2. Observed default rate by marital status.

The gap between unmarried and widow or widower customers is approximately 3.2 percentage points. That difference is worth investigating, but this analysis cannot show that marital status itself causes default. Age, employment, income and other characteristics may overlap with these categories.

Higher income does not create a perfectly smooth risk gradient

The upper-middle and lower-middle groups record the highest observed default rates at 8.7% and 8.6%. The low-income group records 8.0%, while the high-income group has the lowest rate at 7.1%.

Figure 3. Observed default rate by income category.

The total spread is modest—about 1.6 percentage points. These income categories are quartiles calculated from this dataset, so they describe relative position within the sample rather than fixed monetary thresholds that can be reused in another portfolio.

Loan purpose produces one of the clearest differences

Car loans have the highest observed default rate at 9.4%, closely followed by education loans at 9.2%. Wedding loans record 8.0%, while real-estate loans have the lowest rate at 7.2%.

Figure 4. Observed default rate by loan purpose.

The difference between car and real-estate loans is approximately 2.1 percentage points, and both groups contain thousands of customers. That makes loan purpose a useful candidate for deeper modelling, although it still should not be treated as a stand-alone cause of default.

What the bank should do next

·        Calculate confidence intervals or run significance tests to distinguish stable differences from random variation.

·        Build a multivariate model that controls for income, age, employment type, household profile and loan purpose simultaneously.

·        Validate model performance on data that was not used to develop the model.

·        Review the use of demographic characteristics against applicable fair-lending, discrimination and privacy requirements.

·        Monitor default rates over time to determine whether the relationships remain stable.

Methodology in brief

The analysis removed duplicate records, standardised inconsistent text categories, corrected implausible child-count and employment-duration entries, and restored missing income and employment values using age-group medians. Loan descriptions were consolidated into car, education, real-estate and wedding categories. Default rate was calculated as the share of customers in each group whose debt indicator showed a historical default.

Limitations

This is a descriptive, observational analysis. It identifies associations but does not prove causation. No confidence intervals, hypothesis tests or multivariate controls were applied. Very small groups—especially borrowers with four or five children—should not be used for policy decisions without additional data.

Conclusion

The portfolio shows higher observed default rates among unmarried borrowers and customers borrowing for cars or education, while high-income and real-estate borrowers show lower rates. The relationship with children is uneven. These results are most valuable as signals for further investigation—not as automatic lending rules.

Suggested closing question

Which factor would you examine next: employment type, education level or borrower age? Notebook: https://colab.research.google.com/drive/1E0cFAp-Kr_C10rUHQ0qc875lGC0TDRV0#scrollTo=sNEM7wkr-F7N