In our previous publication, we analyzed the latest arrears developments observed in the Dutch mortgage market. Which went back to pre-pandemic levels in 2020 Q2. In this article, we will continue our analysis on Dutch mortgage arrears and dive deeper into what drives the developments of Dutch mortgage arrears.

A study conducted by the Dutch Central Bank (DNB) back in 2017 (Stanfa, Vlahu, & de Haan, 2017) examined the incidence of mortgage arrears in a large sample of countries and explored the role played by various factors in explaining cross-country and within country differences in delinquency rates. These factors were grouped into four main categories: macroeconomic variables, macro-prudential regulation, institutional factors, and housing market characteristics. Specifically, in regards to macroeconomic factors, unemployment rate, house price index movements and interest rates spreads were used. Using a dataset covering 26 countries over 2000-2014, the authors analyzed cross-country and within-country differences in mortgage arrears and although the main focus of the study was the impact of macro-prudential policies, they also found a statistically significant impact of all macroeconomic drivers used for the analysis. In this analysis, LoanClear uses the same macroecnomic variables as in the DNB study to investigate if these features are also relevant when looking specifically at Dutch mortgage loans. The other factors included in the DNB analysis are not taken into account as the dataset used for the analysis consists only of Dutch mortgages and thus, the macro-predential framework and institutional quality are the same for all loans. In addition to these variables, LoanClear examined other factors not covered by the study mentioned above that can explain arrears developments.

Description of the data

This section describes the various data sources from which information on the macroeconomic variables and arrears have been obtained.

European DataWarehouse (EDW)

To start with, for this analysis, we are making use of the European DataWarehouse (EDW) dataset which contains loan part level data for loans that are part of Dutch ECB eligible RMBS deals. The dataset contains data from RMBS deals from 21 different originators with a total outstanding amount of approximately €115 billion as of 2021 Q2. LoanClear constructed a delinquency index based on the EDW dataset which is used as dependent variable for the regression analysis presented in the next section and measures the ratio of delinquent balances over the total balance of the sample through time.

Macroeconomic Variables

We take into account macroeconomic conditions by using three variables: unemployment, changes in house prices and interest rates spread. Unemployment rate and House Price Index (HPI) data was sourced from the Dutch Central Bureau of Statistics (CBS). Finally, the interest rate spread was calculated as the average interest rate spread of the top 6 lowest mortgage rates offered in the Dutch market at the end of each quarter by using mortgage rates data from Hypotheekbond with the following characteristics: fixed rate periods of 60, 120, 240 and 360, annuity repayment type and LTVs of 60, 80 100 and NHG. The average mortage rates desribed above, were compared to the 3-month EURIBOR swap in order to determine the interest rate spread.

Methodology

In this section, we elaborate on the analysis in which we try to find potential factors affecting mortgage delinquency rates. We focus on the macroeconomic indicators mentioned in the DNB study. We assume a one period lag for all macroeconomic regressors with two objectives in mind: (i) we want to mitigate potential reverse causality between arrears and explanatory variables, and (ii) we want to control for the delayed effect that some of the explanatory variables might have on mortgage arrears.

LoanClear has performed a beta regression analysis on historical data of arrears and the macroeconomic indicators mentioned above as well as the employment status. The reason beta regression was used is that this type of regression models is commonly used to model variables that assume values in the standard unit interval (0, 1).

In particular, in case of a beta regression the model looks as follows:

where the logit(yᵢ) is the log ratio of proportions of delinquent and non-delinquent balances, a₀ the intercept and βᵢ the coefficients of the independent variables xᵢ. Thus the beta coefficients are the additional increase (or decrease if the beta is negative) in the ratio of the expected value of proportions :

Table 1 presents the results of the beta regression by taking into account only the macroeconomic indicators that were introduced by DNB. Our findings suggest that yearly HPI change and interest rate spreads are not statistically significant.

One possible explanation for the HPI variable to not have a significant effect on arrears is the lack of strategic defaults, where the borrower decides to stop making payments because they have negative equity in the collateral, in The Netherlands. This is not an attractive option for a borrower in the Netherlands due to the fact that there is a full recourse to the borrower. Thus the HPI change and its effect on borrower’s equity doesn’t have a significant impact on delinquencies. In regards to interest rate spread, the non-statistically significant impact might be explained by the fact that the majority of the loans are fixed interest rate loans (approximately 88% as of 2021 Q2) and thus, the contractual interest payment is not affected by spread developments for the majority of the market.

Table 1: Beta regression results including only macroeconomic indicators

On the other hand, the unemployment rate seems to be the only statistically significant macroeconomic driver. In particular, an increase in unemployment by one percentage point would result in a relative increase of 21% on the ratio of delinquent – non delinquent balances. The relationship between unemployment and arrears in our sample is graphically illustrated in Figure 1 below.

Figure 1: Development of Dutch mortgage arrears segments.
Source: LoanClear, European DataWarehouse

In Figure 1, we can observe that the delinquency index for 3+ months in arrears follows a similar trend to unemployment rate. This becomes more clear when focusing on periods of increasing unemployment such as 2013 Q1 and 2020 Q2. Especially in 2013, unemployment increased from 6.8% in the first quarter to 7.1% during the second quarter. However, the delinquency index for 3+ months in arrears started to increase only one quarter later in Q3 implying the delayed impact of unemployment on arrears. The same pattern though is not as clearly observed for the 1 month delinquency index. This segment is typically more volatile because it’s more common for borrowers to miss one payment due to a temporary lack of liquidity and then repay the arrears balance immediately once they are informed about the missed payment.

In addition to the macroeconomic indicators used by DNB, LoanClear examined the impact of the various employment types and debt service ratio buckets on the delinquency index.

The results can be seen in Table 2. As someone would expect, Self-employed borrowers are the most likely to become delinquent with an impact on the ratio of delinquent – non delinquent balances of 1.58.

Table 2: Beta regression results including unemployment, employment type and debt service ratio buckets

Figure 2: Development of Dutch mortgage arrears index by employment segments.
Source: LoanClear, European DataWarehouse

In Figure 2, the total arrears index by employment type is visualized. Two main observations can be made: i) there seems to be a consistent difference in arrears levels for different type of employment statusses, with arrears among self-employed borrowers being highest and for pensioners being the lowest, and ii) Self-employed borrowers seems to be affected more during an economic downturn whereas almost no effect is shown for pensioners.

In addition to employment type, LoanClear grouped borrowers by Debt Service Coverage Ratio Buckets. The debt service coverage ratio for each borrower was calculated as the monthly gross income divided by the payment due at origination. Based on this calculation, 3 buckets were created: Bucket 1 for borrowers with debt service coverage ratio <= 4, Bucket 2 with a debt service coverage ratio between 4 and 6 and finally Bucket 3 for the remaining borrowers. As can be seen from Table 2, the higher the debt service coverage ratio bucket that a borrower belongs to, the lower the ratio of delinquent balances over non-delinquent balances becomes. In other words, the relative size of income over monthly mortgage payment affects significantly the arrears development as the higher that ratio is, the less likely it is that a borrower will be in arrears. It is also worth mentioning that the percentage of loans belonging to Bucket 1 has been decreasing since 2013. This decrease is probably caused by the tightening of underwriting criteria over the past years.

Conclusion

Using the EDW dataset as well as macroeconomic indicators, we examine potential drivers of arrears in the Dutch mortgage market. Our starting point is the DNB analysis based on which we investigate the impact of unemployment, HPI changes and interest rate spreads on mortgage arrears. Our results suggest that unemployment is the only statistically significant macroeconomic driver of arrears as it directly affects the capacity of a borrower to meet his monthly obligations. In addition to unemployment, employment status and the debt service coverage ratio were introduced in the analysis. The results indicate that self-employed borrowers tend to be the most prone to arrears compared to pensioners and full employed borrowers. Finally, the debt service coverage ratio is found to have a statistically significant impact. It is likely that the increasing proportion of higher debt service coverage ratios observed over the last years is triggered by the stricter underwriting criteria which follow the ‘Tijdeliijke regeling hypothecair krediet’ legislation and is present since 2013.

Bibliography

Stanfa, I., Vlahu, R., & de Haan, J. (2017). Mortgage arrears, regulation and institutions: Cross-country evidence. DNB.

Whitley, J., Windram, R., & Cox, P. (2004). An Empirical Model of Household Arrears. SSRN Electronic Journal.