Credit costs, NPLs to rise, says Maybank Kim Eng’s AI algorithm

On Sept 12, Maybank Kim Eng (MBKE) said in a report that it was using artificial intelligence to gauge the different metrics affecting the non-performing loans (NPLs) of local banks. According to its analysis, the pace of non-Singapore dollar loan growth, domestic inflation and the rate of change in special-mention loans have the strongest influence in setting the direction of NPLs. In contrast, interest rates, changes to unemployment and GDP growth — variables that traditionally guide NPL forecasts — seem to have a smaller impact.

Singapore has a small domestic economy and the local banks have regionalised to maintain earnings growth. No surprise, then, that banks’ earnings growth has outperformed the GDP growth of the Singapore economy. Bank CEOs have also attributed their earnings growth to Singapore’s “hub” economy. Many operations of MNCs and Asian corporates that operate in Southeast Asia are based in Singapore. They include the wealth management activities of private banks and fund management activities of large fund houses.

Most of the banks’ activities are not directly related to Singapore but to the hub status, and that allows banks to deliver higher growth than in the domestic market, Samuel Tsien, CEO of Oversea-Chinese Banking Corp said on Aug 2 during the bank’s 1HFY2020 results briefing.

OCBC and United Overseas Bank have built up regional franchises. The former’s focus is on expanding in North Asia with its Greater Bay Area strategy. UOB is expanding in the Asean region. Its branchless digital bank TMRW was launched in Thailand in March, and a second Asean country is likely to be added by year-end.

According to MBKE, there is a downside to geographical expansion. “With overseas lending accounting for a large share of loan growth in the past three years, NPL risks may height

en as the cycle turns,” the research house says.

By using supervised machine learning algorithms similar to those employed by social-media and video-streaming services for recommendations, MBKE was able to identify variables that have the highest influence on banking NPLs. Its results show that ACU (Asian currency unit) or non-Singapore dollar loan growth — particularly business loans — is a key indicator of NPL growth. The consumer price index, changes in special-mention loans as well as changes in property prices also show correlations to growth in NPL formation, MBKE says. It adds, however, that changes in GDP, unemployment and changes in the Singapore interbank offered rate (Sibor) have weaker correlations.

MBKE estimates that, in the past three years, 63% of incremental loan growth for Singapore banks was fuelled by overseas lending. This implies that increased non-Singapore dollar exposure raises asset-quality risks.

According to 1HFY2019 results, 52% of UOB’s loans are from Singapore, and 41% of OCBC’s loans are from Singapore. For both banks, customer loans by geography are based on where the credit risks reside, which may be different from the borrower’s country of residence or the booking location of the loans. Singapore loans account for 62.8% of DBS Group Holdings’ loan book.

“Combining aggressive overseas loan exposure in the past three years and the results of our machine learning analysis, we believe Singapore banks’ overseas exposure will be the main source of their NPL growth. As a result, even as we forecast credit charges of 25 basis points for 2019 and 29bps for 2020, up from 16bps in 2018, risks are likely to be on the upside,” MBKE says.

Still, the banks’ share prices are likely to be supported by strong capital positions — the three banks have common equity tier-1 ratios above 13.5% and visible dividends, with yields above 4% (see table). E