The Novantas Perspective series provides timely and expert viewpoints on a variety of detailed banking industry subjects.
It finally happened: this year’s CCAR Severely Adverse scenario has a “severe global recession” with a negative three-month Treasury rate. Since the U.S. has never experienced negative rates, this scenario is in the “out of sample” range of every bank’s Stress Testing model — meaning that non-modeled approaches will play a bigger role in this year’s submission. In this Perspective, we provide our first-blush expectation for (1) what would happen in this Severely Adverse scenario and (2) how it affects PPNR modeling. On the former, we believe that negative short-term rates would further erode NIM but would not (at least under the Fed scenario) cause the “end of banking as we know it.” On the latter, PPNR rate models will need to be reviewed for logical outcomes given negative rates, as will most PPNR balance models depending on specification; the process to create an “extended variable” set may have embedded risk that also must be well understood.
Executive Cheat Sheet:
- The 2016 Severely Adverse CCAR scenario has a severe global recession that drives 3m Treasury rates to –50 bp for the scenario duration — but “no additional financial market disruptions” from negative rates.
- The U.S. has never experienced negative interest rates, which makes this an “out of sample” scenario; the best available comparison set are other developed economies with negative rates.
- Switzerland, Sweden, and Denmark have held negative rates since 2012: Banks experienced significant NIM compression; lending rates there fell, causing moderate loan growth; deposit rates, except for large institutional depositors, have not gone negative.
- The UK central bank has considered the influence of negative interest rates, and Japan’s central bank just decided on Friday to charge 10 bp on deposit balances in excess of reserve. Neither country expects an outsized reaction to marginally negative rates.
- While there may be quantum behavior changes from negative rates, the Fed’s scenario explicitly assumes away this and other financial market disruptions.
- Some categories of PPNR Stress Testing models warrant immediate scrutiny, in particular all rate models, but also balance models depending on their specifications.
- Separately, beware “second order” variables that expand the independent variable data sets, as the “out of sample” risk there could be magnified.
Severely Adverse Scenario Specifics
Here is key Severely Adverse language from the Fed’s supervisory scenarios released last week.1
“The Severely Adverse scenario is characterized by a severe global recession, accompanied by a period of heightened corporate financial stress and negative yields for short-term U.S. Treasury securities…. As a result of the decline in real activity and subdued inflation, short-term Treasury rate fall to negative ½ percent [-50 bp] by mid-2016 and remain at that level through the end of the scenario. For the purposes of this scenario, it is assumed that the adjustment to negative short-term interest rates proceeds with no additional financial market disruptions [emphasis added].”
Other macroeconomic factors follow similar trends to the 2008-2009 crisis and are more extreme than the 2015 Severely Adverse Scenario: GDP contracts, equities lose ~50% of value, housing markets collapse (particularly in areas experiencing recent growth), commercial loan markets dry up, and unemployment returns to 10%. Mature markets in Europe, the UK, and Japan take similarly serious blows while emerging markets are hit but not as hard.
Global Experience with Negative Interest Rates
While U.S. Treasuries have traded at 0 bp with increasing frequency, the U.S. has never seen a market environment with persistently negative interest rates.2 In the absence of direct American experience with negative interest rates, the most applicable comparisons are to smaller developed economies that have experienced (Switzerland, Sweden, Denmark, recently Japan) or considered (Great Britain) negative interest rates. The experience can be broadly summarized as follows:
- Deposit rates, as expected, tend to move lower. Consumer deposit rates remain at or above zero; Commercial deposit rates are more likely to be negative. Accounts “earning” negative rates see some outflow, but there have been no bank runs.3
- Lending rates, as expected, also decline. As a rule, cheaper borrowing has yielded a small uptick in Consumer borrowing4 and a larger uptick in Commercial lending.5
- Bank reluctance to take deposit rates negative, paired with decreasing lending rates, have resulted in margin compression that is not offset by increases in lending volume.
- There has been little observed quantum changes in holding cash as opposed to deposits, reflecting the cost associated with holding cash.6
Major economy central banks, while not having experienced negative rates, have considered their implications. Notably, Charles Bean, Deputy Governor of Monetary Policy for the Bank of England, published a paper on the Bank’s expectation if rates went negative. The UK paper found that retail rates were generally not expected to go into negative territory, while wholesale rates would need to drop below –50 bp for at least a year or two before businesses would begin to consider the investments needed for secure holding of cash. It also expected NIM compression and a reduction in profitability. One of the policies recommended matches what the Japanese central bank just did, which is to impose a 10 bp charge on bank funds held in excess of reserve requirement.7
Impact of Severely Adverse Scenario on U.S. Banking
We do not expect the –50 bp rate for 3m Treasuries in the Severely Adverse scenario (with longer-term rates maintaining 0 bp or positive values) to generate material discontinuity between customers’ past behavior and future behavior — first and foremost because the Severely Adverse scenario’s own description precludes such a reaction. That is also consistent with conventional wisdom from international Central Banks, which expect such a reaction only from aggressively negative and persistent rates, i.e., more than –50 bp. It also reflects the strong likelihood that U.S. banks will follow European banks in not taking Consumer deposit rates negative.
We do expect the interest rate forecast to affect bank rates and NIMs, as well as bank balances. In terms of bank rates:
- Deposit rates will come down further, but banks will keep rates positive regardless of algorithmic forecasts which could take rates negative.
- Products with rates tied to market rates may result in unacceptable rates; in these cases, banks may discontinue products or redesign interest rate and fee policies to minimize their impact.
- Loan rates will also decline as will NIM; banks with more rate sensitive balance sheets will suffer more than others, exactly the same banks that are positioned to gain from rising rates today.
In terms of the impact of this Severely Adverse scenario on balances, we generally expect to see similar balance evolution trends to what was witnessed during the last financial crisis:
Consumer deposits would likely grow:
- Affluent consumers — who represent the lion’s share of consumer deposits — will exhibit a “flight to safety” as equity markets contract, increasing their deposits.
- Lower income consumers will be hit hard, especially as unemployment rises, leading to lower deposits and increased overdraft — but representing a small fraction of balances.
- Consumer lending growth would slow or go negative:
- In negative interest rate countries thus far, consumers have shown modest borrowing growth in housing; but in the U.S. the scenario’s instructions call for the housing market to dry up, which reflects consumer caution toward higher housing debt.
- Very low lending rates may move a greater share of household operating funds from deposits to credit card revolving balances; however, the Severely Adverse guidance for tightening risk policies may also dampen this asset class.
Commercial is a mixed bag:
- In negative rate countries, wholesale deposit rates went negative while loan rates were very low; a similar experience in the U.S. would reduce non-operational commercial deposits and generate incremental commercial lending (which would also favorably affect LCR calculations).
- But the Scenario guidance is explicit that CRE and other lending vehicles dry up; absence of credit in the 2008 crisis led many businesses to develop their own “fortress balance sheets” by reducing their reliance on lending and instead bolstering their own deposit books.
We strongly believe that a 3+ year timeframe of negative rates will drive at least one large U.S. bank to test negative deposit rates. However, we strongly advise not to rely on this thought experiment in shaping expectations, since the size of this test would likely be small, the timing of the test would likely be towards the end of the ~3 year Severely Adverse scenario, and predicting the learning from such a test is problematic.
Impact of Severely Adverse Scenario on CCAR Stress Testing
Negative interest rates represent an “out of sample” forecast, where historic experience is no longer representative of the question being asked. That means that the outcomes from existing CCAR rate and balance models will be inherently suspect for use in predicting outcomes for this Severely Adverse scenario.
As a single example, we used one of our off-the-shelf test models to predict industry-wide liquid savings rates. Given a purely algorithmic view, the model forecasts that a savings account would “earn” –20 bp in 2017. However, there is every reason to believe that banks will not “cross-0” on their consumer deposits, obviating confidence in the model results.
The “out of sample” problem is mitigated — but is still a risk — given our conceptual conclusion and scenario guidance that customers will likely view this as simply another rate reduction and not a catastrophic break in the financial system. Nevertheless, the validity of the model rests on not only whether the reaction curve is smooth and continuous, but also whether the model can properly be applied in the range where there is no historical data.
There are models and non-modeled approaches that, given their structures and the line items being predicted, are likely to have a higher risk of failing when interest rates turn negative.
Below are the families of models that are of the most concern to us:
- All rate models, with the highest likelihood of failure being any deposit model that forecasts negative product portfolio rates
- Deposit balance models that predict off of either short-term or long-term rates without considering the difference between the two.
- CD balance models, especially in later years when the 10y Treasury is materially higher than the –50 bp 3m Treasury; e.g.; this includes acquisitions by tenure, CD Renewals, and term reassignment estimates.
- Balance mix models (e.g., Consumer MMDA vs. CD; interest-bearing vs. non-interest bearing), especially those with relatively basic conceptual structures.
- Models for “benchmark-plus” products where product rates are anchored on wholesale rates, especially if product rate can go negative.
- Mortgage, CRE, or other lending models that are over-reliant on rate sensitivity and under-reliant on macroeconomic sensitivity (the opposite is also a problem, though less frequent in our experience).
We also strongly advise that banks review the process by which any “second-order” variables are created. Many banks expand their variable set, and we expect these macroeconomic conditions to be significantly more sensitive to “out of sample” risks. As such, the forecasts for these variables can either be:
- Algorithmically generated without material judgmental review or consideration of extreme scenarios; in this case are you confident enough in these variables to appropriately account for the momentously out-of-sample realities of negative interest rates?
- Based predominantly on intuition, assumptions, and business judgment; in this case, do you know who is driving those assumptions and what decisions they are making?
Finally, the uniqueness and oddity of this scenario is a clear example why it is important to have a robust and forward-thinking plan to develop non-modeled approaches and overlays. We expect we will be seeing a lot of both for stress testing this Severely Adverse scenario to reflect appropriately how customers and bank management will react.
- Available here: http://www.federalreserve.gov/newsevents/press/bcreg/bcreg20160128a2.pdf
- See details on the Treasury website: https://www.treasury.gov/resource-center/data-chart-center/interest-rates/Pages/TextView.aspx?data=billrates
- “Less than Zero: Living with Negative Interest Rates”, Wall Street Journal, updated 12/8/2015 http://www.wsj.com/articles/less-than-zero-living-with-negative-rates-1449621094
- Swiss National Bank Financial Stability Report 2015 available at https://www.snb.ch/en/mmr/reference/stabrep_2015/source/stabrep_2015.en.pdf
- “Bankers vs Mattresses”, Economist, updated 11/28/2015 http://www.economist.com/news/finance-and-economics/21679231-central-banks-are-still-testing-limits-how-low-interest-rates-can-go-bankers
- “How Far Can the Repo Rate be Cut?”, Riksbank, 9/30/2015 http://www.riksbank.se/Documents/Rapporter/Ekonomiska_kommentarer/2015/rap_ek_kom_nr11_150929_eng.pdf
- Available here: http://www.bankofengland.co.uk/publications/Documents/other/treasurycommittee/ir/tsc160513.pdf