Italian banks hold a large amount of debt that is at or near default, and many investors are alarmed. Yet at current levels, we believe the Italian non-performing loan (NPL) problem could be managed over time through a combination of earnings growth and collateral recovery. However, with market uncertainty spiking post-Brexit, policymakers now feel pressured to pursue a quicker resolution.
Under the current framework of the European Union Bank Recovery and Resolution Directive (BBRD), a “bail-in” would impose costs on investors, including the Italian taxpayers who represent a significant portion of them. Consequently, while acknowledging the difficulties of this under BBRD, we think that one option to consider is the Italian government buying the NPLs. It could be funded at the government borrowing rate, and based on current prices and potential recoveries, could potentially be profitable fortaxpayers in the long run. This can be thought of as a form of government investment rather than state aid. Given the fluidity of the situation, we ultimately expect some combination of the various resolution options to be employed, consistent with the BBRD framework.
How big is the current problem?
Non-performing loans vs. non-performing exposures
A lot of uncertainty right now stems from clarifying the size of the problem. The worst-performing loans in Italy are known as sofferenze, which can be broadly translated as “very bad debt.” It refers to those loans that have already defaulted. It is this category, which is currently €87 billion on a net basis,1 that the banks and government consider the stock of NPLs.
Many regulators and market participants, however, prefer to look at non-performing exposures (NPEs). As well as the stock of NPLs, this includes loans unlikely to pay in the future, or that are currently past due, but are not yet bad enough to classify as NPLs. While transparency around loan performance could be improved, our understanding is that loans over 120 days past due are classified as NPEs at a minimum.
The question for investors – and the reason for much uncertainty today – is whether NPEs are actually NPLs, and whether the banking sector’s exposure is closer to €87 billion or €197 billion? Markets and regulators appear to be saying yes, while banks and governments are saying no. From our perspective, we think the €197 billion of net NPEs is the best place to start. Of this, around 60% is concentrated in five banks (see Figure 1).
Looking ahead, it’s possible that loan quality could deteriorate further. One positive sign is that the growth of NPLs has slowed markedly (see Figure 2), suggesting some level of “burnout” consistent with historical norms. However, this could change if macroeconomic conditions weaken; the challenge for investors is knowing where we are in the credit cycle. Using transition rates (the portion of performing loans converting to non-performing loans) from before the financial crisis paints a very benign picture of future exposure, while using recent rates paints a more challenging one.
What does this mean for the banking system?
System-wide losses on NPLs look manageable over time
In our view, the weakest banks in Italy will likely need to raise capital in the short term - a process that is well underway. However, assuming the situation doesn’t significantly deteriorate, the NPL exposure looks manageable for the banking system in aggregate over time, for three reasons:
- Expected earnings: System-wide pre-provision earnings (PPE)2 are currently ~€30 billion per annum. At this level, banks can write down half of all their current net NPEs to zero, with zero recovery, over the next three years, with limited impact on capital levels. However, the key assumptions are that Italy avoids a significant recession, and current earnings are sustainable. The policy objective of making the system recession proof today, means that earnings are being given minimal, if any, credit by market participants right now.
- Collateral recoveries: Data is not easily available, but a Moody’s report published in 2014 estimates loss severities of 40% on secured NPLs and 80% on unsecured NPLs. According to a PwC report from June 2016, 53% of gross Italian NPL is secured, suggesting the amount of net NPL is largely covered by real estate collateral.3
In addition, a certain portion of the loans have personal guarantees, which could provide higher collateral recovery over time. While it is difficult to assess this number accurately, PwC estimates that 19% of the NPLs are retail loans. Acknowledging that the better loans are probably securitized, and that there are idiosyncrasies in Italy that may make recoveries harder than in countries such as Spain and Ireland (for example, the long resolution process through the courts, which on average can take six to seven years), these numbers suggest that the current marks are reasonable over the long term.
- Strategic defaulting: Although this is extremely hard to quantify, our research suggests that strategic defaulting – where a borrower temporarily chooses not to pay, but can ultimately do so – is prevalent across Italy. This is due to the time it takes to foreclose on a loan, which is often greater than five years. If strategic defaulting is prevalent it means that end losses to banks will be less than the current NPE number implies. As such we think providing greater transparency on the extent of strategic defaulting would go a long way toward improving market confidence. In addition, adopting a zero tolerance policy towards strategic defaulting in the first place, in combination with shortening foreclosure times, could help maximize collateral recovery.
Are current market prices for the non-performing loans fair?
Concerns on mispricing look overstated
A key concern for investors is that the NPE and NPLs are mismarked on bank balance sheets, meaning that banks should be writing down higher exposures than they are currently doing. In our view this is debatable. As argued above, if the NPEs are held to maturity, we think that the banking system is adequately covered in aggregate, when considering collateral, guarantees and strategic defaulting.
However, it’s also fair to say that if the NPLs were sold under current market conditions, there would likely be a 10 to 20 percentage point difference between where investors would bid to buy the loans and where they have most recently been valued. (For example, an NPL currently valued by a bank at 40 cents would command between 20 and 30 cents if sold in the market today.) To sell all the current NPLs in the system at these market prices would lead to a capital depletion of around €40 billion. Although this seems large, it equates to just over one year of system-wide PPE. To be clear, this is a system-wide number and doesn’t negate the need for the weaker banks to potentially raise capital.
Differences between bid and book values are driven by funding rates, asset value and expense assumptions
This difference between the bid (the amount people are willing to buy for in the market) and the book value (the level at which banks currently hold the loans) is driven by the following factors:
- Funding (discount) rates: The liability structure of the bank selling the NPLs is likely more efficient than the potential purchaser of the NPLs. For example, the cost of funding for banks, based on figures from the Italian Banking Association, is 1.08%, including deposits and retail bonds. A potential bidder, such as a hedge fund or private equity firm, will be challenged to get those funding rates. Given that Italian NPLs are a longer-duration asset, these funding differentials can be meaningful. The new GAC program (Garanzia Cartolarizzazione Sofferenze, or NPL Securitization Guarantee) should narrow the bid-offer by several points, but likely doesn’t lower the funding costs enough to justify paying substantially higher for the NPLs.
- Asset value assumptions: Residual asset value assumptions (how much of the loan will ultimately be recovered) can differ significantly depending on the investor type, given differences in assumptions about Italian macroeconomic dynamics, the recoverability of collateral and liquidation timelines.
- Expense assumptions: Most banks expense the servicing/workout cost of NPLs when the expense occurs (the high level of NPLs is one of the reasons Italian banks have higher expense ratios). However, the NPL purchaser discounts these expenses at the purchase discount rate. This again leads to the purchaser seeking a lower price.
Given these differences, Italian banks look to be rational in holding on to their non-performing loans. It’s also important to note that most banks would be challenged to sell their illiquid assets at current marks.
What can be done to resolve the problem?
There are three main options for reducing NPLs across the Italian banking sector, all of which have their pros and cons.
- Bail-in: The BRRD framework provides for recapitalization of banks, but imposes burden sharing on investors, or impairment of the bank’s liabilities. Conceptually, the positives of this approach are that it would clean up the system quickly, facilitate lending to the wider economy and provide better protection should another recession occur. However, there could also be negative consequences, including the potential of triggering a wider systemic shock. As we’ve cautioned before, this could prove counter-cyclical, raising the cost of equity and borrowing on weaker sovereigns and lower income borrowers. It would also impose an immediate cost on Italian retail investors, who constitute around 50% of subordinated debt holders.
- Muddle through: An alternative option would be for the sector to muddle through, addressing the weakest lenders as needed, and using earnings to write down exposures over time. This mitigates the risk of a systemic shock and means that investors do not need to take immediate losses and taxpayers do not need to provide a subsidy. However, the NPLs remain as an overhang in the system, limiting lending and future flexibility in the event of a recession. That said, thus far, there is limited evidence that recapitalizing banks has led to extensive lending particularly to those most in need.
- Recapitalization using low-cost funding: A final option, albeit one that is significantly harder post-BBRD, is to remove the NPLs from bank balance sheets using low-cost funding. As with a bail-in, the pros of this approach are that it cleans up the system quickly and facilitates future lending – albeit with the cost potentially borne by taxpayers, not investors. Although this may seem unpalatable, we think it potentially could be positive for taxpayers: Given low funding costs, if the government were to purchase secured NPLs, a taxpayer could potentially earn in excess of a double-digit return on select secured loans purchased at current book value, assuming a 60% loan-to-cost funding at the government funding rates. Said another way, our analysis shows that the price of an unlevered secured pool, trading with an expected IRR in the low to mid-teens range, could potentially be improved by 10 to 20percentage points, with 1% funding at a 60% loan-to-cost ratio.
Conceptually, this is not too dissimilar from governments investing in other assets, be it infrastructure funds or private companies. One key advantage is that it allows taxpayers, rather than private investors, to benefit if the economy recovers. It can therefore be thought of as a form of government investment rather than state aid.
Overall, the Italian NPL problem is a manageable one that could be addressed with a muddle-through approach. Before the Brexit vote, this was arguably the best solution. However, the uncertainty following Brexit, and concerns about contagion risk, mean that a quicker resolution may now be required. While acknowledging that this is challenging under the current BBRD framework, we think that over time the system should differentiate between bail-ins, where investors take losses, and funded solutions that could potentially be profitable for the taxpayer in the long run. In the case of Italy, where the investors are also the taxpayers, the latter could be a potentially elegant way of resolving the problem. It would not impose immediate losses on Italian retail bond investors and would preserve the option of limiting or eliminating potential tax payer losses over time. Given the fluidity of the situation, we ultimately expect some combination of the various options to be employed, consistent with the BRRD framework.
1Net loan value is the gross value of loans minus provisions already set aside for defaults
2Earnings before provisions for future loan defaults are taken into account
3“The Italian NPL Market: The NPL Volcano is Ready To Erupt,” PwC, June 2016