UBS Survey of recent mortgage holders

Work Undertaken

Between 7th of July and 4th of August 2017 UBS Evidence Lab conducted an online survey of 907 Australians who had recently taken out a mortgage to buy a residential property. This follows on from our August/September 2016 survey which covered both the 2015 and 2016 Vintages. Respondents were required to be personally and deeply involved in the discussion and completion of the mortgage paperwork. Respondents surveyed were aged 18 or above.

This detailed survey asked 70 questions around the borrowers’ background, motivations, purchase method and expectations. The survey was broad based, covering all states and territories in Australia. Given the size of the sample and broad spread of respondents we believe the results are representative of Australian mortgage borrowers. Conclusions based on the total 2017 sample have a potential sampling error of just +/-3.18% at a 95% confidence level. The survey has been split into three cohorts:

  • 2017 Vintage of 907 borrowers who have taken out a residential mortgage during the twelve months to August 2017;
  • 2016 Vintage of 854 borrowers who have taken out a residential mortgage during the twelve months to August 2016;
  • 2015 Vintage of 374 borrowers who have taken out a residential mortgage during the twelve months to August 2015.

Are bank customers telling the truth?

One of the key areas of focus of the 2017 UBS Evidence Lab Australian Mortgages survey was to assess the level of factual accuracy in mortgagor’s applications. While there has been anecdotal evidence for many years that customers are not always accurate in their application, we were lacking hard evidence. As a result the UBS Evidence Lab asked participants who had recently taken out a mortgage the degree of factual accuracy in their application.

The results of this survey were disappointing, with only 67% of participants stating their mortgage application was “completely factual and accurate”. This is a statistically significant fall from the results of the 2015 and 2016 Vintages.

This was offset by a statistically significant increase in respondents who stated their application was “mostly factual and accurate” (25% up from 21% in the 2016 Vintage) and “partially factual and accurate” which reached 8% of applications (up from 6% in the 2016 Vintage and 3% in the 2015 Vintage).

We see these results as disturbing and difficult to reject given approximately onethird of participants stated their application was not entirely factual and accurate.

Further, it is highly unlikely respondents would have stated that they misrepresented their mortgage application when in fact they were truthful. If anything, we believe it is more likely these figures may understate the level of misrepresentation in mortgage applications as some respondents may not want to state they were less than completely accurate despite the anonymity of this survey.

Misrepresentation more prevalent through the broker channel

During the 2017 Survey we found a statistically significantly higher level of factual inaccuracy via the broker channel than via the bank’s proprietary networks.

However, the level of factually inaccuracy has risen across both channels. In the 2017 Vintage only 61% of participants who undertook broker originated mortgages stated they were completely factual and accurate. This is down from 68% in 2016. This compares to 75% of customers stating they were factually accurate via the bank’s proprietary networks. Of concern 11% of participants who took out a mortgage via the broker channel in 2017 stated their application was only “partially factual and accurate”. This is a statistically significant increase from both the 2015 Vintage (4%) and 2016 Vintage (7%).

Was misrepresentation suggested by the broker or banker?

While the significant level of mortgage misrepresentation is a concern, we are more concerned that a substantial number of applicants continue to state that their mortgage consultant suggested they misrepresent their documentation.

During the 2017 Vintage 37% of respondents who used the broker channel and were not completely factual and accurate stated that their mortgage broker suggested they misrepresent the application. While this is down slightly from 41% in the 2016 Vintage it is still a concern.

This also implies that 14% of all mortgage applications via the broker channel were factually inaccurate following the suggestion of their broker, a similar level to 2016 (13%). It is possible that customers who have knowingly misrepresented their loans would prefer to point the blame at someone else, even in an anonymous survey.

While there has also been a slight (not statistically significant) increase in the level of mortgage inaccuracy via the banks’ proprietary networks, the number of participants who stated that their banker suggest they misrepresent has fallen to 8% in 2017 from 13% in 2016. This implies only around 2% of bank originated mortgage applications are inaccurate as a result of the suggestion of the banker.

Factual accuracy at an individual bank level

In assessing the level of factual accuracy of mortgage applications we also benchmarked across the banks (via both proprietary and broker channels). While there was no statistically significant difference in mortgage application accuracy in the 2015 or 2016 Vintages a divergence has occurred in 2017. In the 2017 Vintage Survey:

  • 55% of respondents who took out a mortgage with ANZ (via both proprietary and broker channels) stated their application was completely factual and accurate (implying 45% of customers misstated their application) down from 66% in 2016.
  • ANZ’s level of mortgage application factual accuracy (55%) was statistically significantly lower than the Industry (67%) for the 2017 Vintage at both a 95% and 99% confidence level.
  • NAB saw a statistically significant fall in “completely factual and accurate” mortgage applications in the 2017 Vintage to 62%. However, factual accuracy at NAB was not statistically different to the Industry.
  • While the survey shows a deterioration in the level of factual accuracy at CBA over the last three years, this deterioration was not statistically significant, nor was the difference between CBA customer responses and the survey total.

We undertook further tests to assess potential explanations for the increase in mortgage application misstatement at an individual bank level. However, at a bank level there were no statistically significant differences between any bank and the Industry as a result of origination channel (broker vs proprietary), state, purpose of mortgage or other factor which could easily explain the deterioration in underwriting.

In what way did applicants misrepresent their applications?

Survey participants who did not select they were “completely factual and accurate” in their application were also asked a follow up question about what aspect they misrepresented in their application. We found:

  • 15% of respondents who were not factually accurate in their application stated that they over-represented their household income. This percentage was similar to previous years.
  • 16% over-declared other assets and 15% under-represented other loans or commitments. These percentages are also broadly flat.
  • 30% of respondents stated they under-represented their living costs. This is up from 27% in the 2016 Vintage and 23% in the 2015 Vintage, although these changes are not statistically significant.
  • Interestingly the largest change was a fall in respondents who were not prepared to answer the question.

When we compare the responses to this question across respondents who took out a mortgage via the broker or proprietary networks we found:

  • 18% of respondents who misrepresented their mortgage applications via the broker network stated they over-represented their household income. This was statistically significantly higher than via the bank network at 6%. (These numbers were broadly consistent with prior vintages);
  • 19% of broker customers who misrepresented their applications stated they over-declared other assets compared to 9% via the branch network.
  • The percentage of customers who understated living expenses was consistent across distribution channels.

The findings from 2017 appear to be consistent with previous vintages, where there was a statistically significantly higher amount of misrepresenting household income via the broker channel compared to the banker channel. However in previous vintages there was no statistically significant difference between broker and banker channels for over-representing other assets.

Extent of over/under representation on mortgage applications

When we conducted our first UBS Evidence Lab survey of Australian mortgage borrowers last year (covering the 2015 and 2016 Vintages) we were surprised not only at the level of misstatement, but also at the preparedness of respondents to answer these questions under anonymous conditions. As a result we added a further question this year asking respondents the extent to which they over/understated elements of their mortgage applications.

We found the median response to our questioning around the extent of over or under-representation on mortgage applications was in the range of 10-12%.

However, there was a wider range of responses for both the level of underrepresentation of other loans or commitments and under-representation of living expenses with some respondents stating they under-stated on their mortgage application by around 30%.

Although we were not surprised by the extent of misrepresentation illustrated in these responses – which is consistent with anecdotal evidence – these findings are nevertheless disturbing, especially as these customers still had their applications approved.

Further, our findings are consistent with the ASIC Review of Mortgage Broker Remuneration which stated:

“We identified significant numbers of loans across several lenders where the consumer expenses were stated to be equal to the HEM benchmark. While lenders and brokers may be able to use benchmarks such as HEM as part of their process for verifying consumers’ expenses, they are still required to make inquiries into the consumer’s actual expenses. The proportion of loans we reviewed where the consumer’s expenses were equal to or very close to the HEM benchmark suggests that these inquiries were not occurring properly.”

Is underwriting being tightened?

A customer’s view One of APRA’s key priorities is sound lending, ensuring the banks have the correct lending protocols in place. This follows a period of looser underwriting standards during the years following the Financial Crisis which led to accelerating levels of household debts and rapidly increasing house prices, especially in the largest cities of Sydney and Melbourne. While the banks have put in place a number of additional processes to comply with APRA’s sound lending guidelines it remains to be seen whether they are effective.

We believe one test of the effectiveness of tighter underwriting practices is to see whether customers are finding it more difficult to attain credit or if customers are required to provide more documentation and evidence before credit is provided.

Ease of loan approval

For the first time in the 2017 Survey we asked participants to rate the ease of attaining a mortgage compared to their previous experiences.

  • 46% of participants stated the loan approval process was either ‘much easier’ or ‘somewhat easier’ than their previous application
  • 35% stated their recent mortgage application was about the same level of difficulty as previous years
  • 17% stated it was somewhat or much more difficult. Given First Home Buyer levels have been near record low levels over the last twelve months this would imply borrowers are not finding it more difficult to attain mortgage credit at all.

It may be argued that these findings are logical as many people who are applying for their second or subsequent mortgage are in a better financial position than they were previously, while the mortgage providers would argue this is a result of better service. However, given the rapid increase in house prices over the last few years the amount of credit (as a multiple of income) required to purchase a property has increased substantially. This would imply that a greater proportion of people should be finding it harder to attain approval.

As the value of mortgage approvals has been broadly stable over the last twelve months and customers are not finding it harder to get approval it is difficult to conclude underwriting standards have been tightened from a customer’s perspective.

It could also be argued that customers are finding it easier to attain credit due to the increased penetration of mortgage brokers, who arguably do much of the application heavy lifting. However, we did not find a meaningful difference in the answer to the question of ease of attaining approval across the proprietary or broker channel.

Further, there did not appear to be a meaningful difference in response to the question of ease of loan approval via vintage. In fact participants suggested ease of attaining approval had improved over every prior vintage back to the 1990s.

Finally, we looked at when in the last twelve months the respondent took out the mortgage. Perhaps underwriting standards were being implemented with a lag? However, when we broke the 2017 Vintage responses into calendar quarters there was no significant difference in response.

Has there been an increase in the amount of mortgage documentation and verification?

As a further test to assess if underwriting standards are in fact being tightened we asked survey respondents whether there had been an increase in the amount of supporting documentation and verification compared to their previous mortgage applications. We found no evidence that there has been an increase in supporting documentation or verification compared with previous mortgages.

  • 27% of respondents stated the level of documentation and verification was either ‘much less’ or ‘somewhat less’ that their previous application;
  • 24% of respondents stated the level of documentation and verification was either ‘much more’ or ‘somewhat more’ than their previous application.

There was no material difference in the response to the level of documentation or verification either by distribution channel (Bank vs Broker) or the level of factual accuracy of the application.

Finally, there was no meaningful difference across vintage of previous application nor the calendar quarter of 2017 when the most recent mortgage was approved. This implies customers who misstated their mortgage application have not seen an increase in the amount of documentation or verification required and were no different to customers who were accurate in their application. We found these responses disappointing. We believe there is little evidence to suggest customers are finding it more difficult to attain credit or that mortgage underwriting standards are being tightened from a customer’s perspective.

$500bn of ‘Liar Loans’ in Australia

During the Financial Crisis the term ‘Liar Loans’ was used in the USA to describe mortgages which were no-doc, low-doc or where the verification was not accurate. In Australia mortgages generally require full documentation (low doc loans represent <1% of new mortgages). However, the fact that approximately one-third of respondents state that their mortgage application was not “completely factual and accurate” suggests they too could be described as ‘Liar loans’.

How big is this problem?

Over the last three vintages the level of ‘completely factual and accurate’ mortgages by survey respondents has fallen from 73% (2015), 72% (2016) to 67% (2017). We do not have accurate data on the level of factual accuracy of mortgage applications prior to this time. However, we believe it is fair to assume that mortgage misstatement has been an ongoing and growing issue in the banking system. This was the reason we initiated this UBS Evidence Lab survey in the first place. Therefore, for the purpose of this analysis we have assumed factual accuracy of around 70-75% for vintages prior to 2015.

In Australia the average life of a mortgage is approximately 4 ½ years. As a result this would imply between ¼ to 1/3 of the mortgages on the banks’ books are based on factually inaccurate applications. This equates to approximately $500bn of factually incorrect mortgages (aka ‘Liar Loans’) on the Australian Banks books or 18% of all outstanding Australian credit.

Does this matter?

When we published our first UBS Evidence Lab mortgage report on the 2015 and 2016 Vintages which showed substantial levels of factual inaccuracy we were frequently asked the questions – So what? Does this really matter? Given this has been an ongoing problem in Australia for some time and mortgage arrears remain at low levels (~70bp) we understand there can be some level of complacency. However, we believe one of the key reasons arrears levels are low is the sustained house price inflation seen in many parts of Australia, especially Sydney and Melbourne. As a result we believe many customers who are under mortgage stress have been able to sell their properties, repay their debts and crystallise a capital gain.

However, in the event that house price inflation stops we believe the problems arising from inaccurate mortgage applications come to the fore.

As seen above, many mortgagors have overstated their income and assets, while others have understated their living expenses and liabilities. In the event that household income growth remains subdued or interest rates continue to rise (via the RBA or bank mortgage repricing) this is likely to place pressure on many households. While this could be offset by reduced spending by leveraged customers this would have a broader economic impact.

Overall, we believe the accelerating level of mortgage misstatement by Australian borrowers is a substantial problem and more needs to be done by the banks to address it. While the banks are attempting to tighten mortgage underwriting standards to comply with APRA’s sound lending practices, there is little evidence to suggest this is working.

As a result we believe both the probability of default and loss in the event of default for the Australian mortgage books continues to be underestimated. Further the impact on the broader economy from a housing downturn is likely to be more severe than the banks currently.

How widespread is the problem?

Over the last two years we have run this survey we have been surprised by the extent of mortgage misrepresentation. When we initially saw the results of the 2015 and 2016 Vintage survey we believed it could be a sample error. However, the 2017 Survey confirmed the extent of mortgage misrepresentation, with the level of factually inaccurate applications rising, not falling (to a 95% confidence level).

We have seen earlier in this report that one of the key determinants of mortgage misrepresentation is the distribution channel, with mortgages coming through the broker channel statistically significantly more likely to misrepresent than via the banks’ proprietary channels. But are there any other factors which may drive misrepresentation?

Implications for the Australian Banks

We believe that 26 consecutive years of GDP growth in Australia has led to a large level of complacency within the economy. This survey suggests many people have come to take house price inflation as a given and are prepared to be factually inaccurate on their mortgage application to ensure they get access to housing leverage. And why not, being long housing has paid off handsomely for decades.

However, as many regulators have highlighted household indebtedness is at concerning levels and housing credit growth is continuing to run at many multiples of household income growth despite macro prudential tightening. The UBS Evidence Lab survey findings suggest that the risks within the mortgage book may be more elevated than the banks believe. Therefore we believe the implications for the banks may be:

  • Loss given default on the mortgage book may be larger than anticipated;
  • The impact of a housing downturn on the broader economy many also be larger than anticipated;
  • More needs to be done by the banks to check documentation and verify mortgage applications. We see requiring ATO tax returns to be provided as a verification of income as an obvious starting point;
  • The risk of mortgage mis-selling litigation may rise if house price inflation stops.

We remain underweight the Australian banks within a regional and global context. We believe there are significant headwinds facing the sector over the medium term.