According to data published by the National Credit Regulator (NCR), South African credit providers originated over R124 billion in mortgage loans in 2013 (data for the full year for 2014 has not yet been published). Roughly 30% of mortgages granted (by number) were for less than R350 000 with 11% of all mortgages (again, by number) going to individuals earning less than R15 000 per month.
According to the NCR, as at end Q3 2014 roughly 3.3% of mortgage loans by number and 3.9% by value were 90 days or more in arrears. The trend is positive, with arrears levels significantly lower than their peak of 6.5% by number and 9.4% by value in 2010. The data published by the SARB for bank lenders tells the same story. However, neither the SARB nor the NCR publish performance data by market segment; their data cannot be used to explore how mortgages granted to lower income households performed relative to those granted to higher income borrowers. This would be useful in order to assess the impact of efforts to expand access to mortgages in previously under-served markets, for both lenders and borrowers.
The best we can do with published data is to explore the relationship between arrears levels measured in terms of number of loans and arrears levels measured by value. The latter has consistently exceeded the former, indicating that loans with smaller balances are less likely to be in arrears compared to larger loans. Interestingly, the gap between these two metrics increased noticeably as performance deteriorated. While it may be tempting to draw the seemingly obvious conclusion that smaller mortgages – typically granted to lower income households – perform better than larger mortgages, this would be erroneous. The data tells us nothing about the size of the mortgage at inception. It tells us only that mortgages that have low balances at this point in time perform relatively well. Presumably, this set of loans is likely to be dominated by older mortgages that are almost paid off, rather than newer low value mortgages. If anything the data highlights how important it is to examine the data by segment rather than simply exploring the average performance of the book as a whole. Sadly, published data cannot facilitate this.
CHART 1: Proportion of loans that are 90 days or more in arrears by value and number
In order to explore performance across segments of the market we have analysed monthly payment data submitted by lenders to credit bureaus. The key challenge with that data is to segment loans across the market. Loan size on its own is not good enough; smaller mortgages may be granted to higher income borrowers who purchase relatively costly homes using equity from the sale of another house. Alternatively higher income borrowers who already have a primary mortgage on their property may take out a smaller secondary mortgage. To overcome this challenge we used two criteria. Firstly we selected only those mortgages granted in affordable areas as defined by CAHF’s Citymark dashboard. In addition, the analysis capped the size of the mortgage to ensure that repayments would be no more than 25% of the upper household income threshold determined annually by the Banking Association to define the so-called affordable housing market given prevailing interest rates.
The analysis tracked the performance of two tranches of loans; those granted between January 2004 and December 2008 under the first phase of the FSC, referred to as FSC loans, and those granted between January 2009 and December 2011 referred to as Affordable Market loans. The charts below summarise the performance of these two tranches of loans relative to the rest of the market. Only five years of payment history is maintained by bureaus. It is therefore not possible to track performance back to loan inception for loans granted early on. Of course, the period of review for the Affordable Market loans is much shorter.
FSC loans have a higher default rate than non FSC loans prior to 2010; with a difference of up to two percentage points visible during some intervals. Thereafter FSC loans track the rest of the market closely. Affordable Market mortgage loans granted between 2009 and 2011 performed significantly better than the FSC loans granted between 2004 and 2008. Once again there is a difference in performance between the Affordable Market loans and the rest of the market, with this difference becoming more significant over time.
Both tranches of loans were granted as fairly dramatic developments played out in the broader economy. In 2006 the interest rate cycle turned. The dominant rate on new mortgages, increased steadily from 12.5% from mid-way through 2006 to 17.5% two years later. By way of example, a household paying R2 000 per month on its mortgage instalment in May 2006 would have had to pay an additional R650 per month two years later. At the same time commodity prices, notably the cost of fuel, increased significantly; motorists spending R450 per month on petrol in May 2006 would have had to spend an additional R267 to purchase the same amount of fuel in June 2008. The cost of living had increased and households entering into a mortgage loan agreement at the threshold of their affordability would have felt this significantly.
With regard to Affordable Market loans, these were originated following the sub-prime mortgage crisis in the USA and the global credit crisis that followed. Mortgage lending slowed dramatically in South Africa; the number of mortgages granted in 2009 was half of that granted in 2008. Banks tightened credit granting processes, reduced loan to value ratios and limited access to equity withdrawal facilities. At the same time interest rates declined sharply, with the dominant rate on new mortgages falling 500 basis points between November 2008 and August 2009 and a further 150 basis point by the end of 2010, settling at a low of 11% in November 2010 and remaining stable for the duration of the period under review.
That there is a difference in performance between the tranches is unsurprising. The surprise perhaps is that the difference is so small. On the one hand this is a significant achievement. Banks should be lauded for providing greater access to mortgages in under-served markets without taking on significantly more risk. As the data clearly shows, mortgages granted to lower income borrowers in South Africa cannot be characterised as subprime. In the USA for instance, delinquency rates on fixed rate subprime mortgages are six times higher than those on prime mortgages. Lenders in South Africa did not behave in any way like their American counterparts, despite political pressure to provide access to mortgages.
On the other hand, one could argue that the data indicates significant scope for further innovation in the mortgage market. There must be significant opportunity to expand access while managing risk sustainably somewhere between the relatively small difference in default rates that we observe across portfolios in South Africa and the unsustainable gap we see in the USA. This is particularly the case in light of the diversity of South Africa’s market of homeowners and potential homeowners. Across the townships and suburbs of the major metros, average values of formally registered residential properties vary widely. For instance, in the City of Cape Town in 2012, average property values in Crossroads were R93 000 compared to over R12.5 million in Clifton just 20 kilometres away. Of course employment patterns and socio economic conditions vary accordingly but the potential mortgage market must surely be able to accommodate more than the one-size-fits-all existing 20 year mortgage products that differ only in size and interest rate.
To support this it is critical that lenders continue to learn from their experience in shifting access downmarket and for lenders and policymakers to have a clear understanding of the key risks and costs that impact on mortgage profitability. While we have explored default rates by broad market segment – and there is far more that could be done in that area if additional data is utilised – this is but one domain of analysis. Areas for further investigation include triggers of default. Why do households default, and how many of these underlying default drivers are insurable events? Mechanisms to manage these risks could protect lenders and borrowers, and enhance access to mortgages without compromising the financial stability of either party. It is also essential to explore lending patterns and performance across areas in South Africa with a particular focus on newly built stock built by the private sector compared to older stock in established townships.Then of course we know dangerously little about foreclosure rates, and to our knowledge no industry-wide study has explored loss given default and how this differs by area and borrower income.
Clearly if lenders are to innovate and move mortgages further down the market, processes and their associated risks and costs in this phase of the mortgage lifecycle area are as important to understand as in any other.
 Citymark has defined affordable areas as those where the average property value is less than R500 000. For more detail on this approach, see http://staging.signpost.co.za/housingfinanceafrica-old/citymark/
 . The essential difference between these two tranches relates to the target market they served and the average loan size involved. For FSC loans the target market was defined as those households earning up to R7 500 per month in 2004 indexed to CPI. In 2008 the upper threshold was R9670. For affordable market loans the income threshold was increased to R15 142 in 2009. In 2011 it was R15 738.
 One or more months in arrears, but not foreclosed
 See http://www.mba.org/NewsandMedia/PressCenter/87218.htm. In the third quarter of 2013 the delinquency rate on fixed rate prime mortgages was 3.23%. It was 19.52% on subprime mortgages