VoxEU Column Financial Markets Financial Regulation and Banking Global governance

Same but different: Comparing non-performing asset measures across countries

Non-performing assets are a double-edged sword. On the one hand, they often trigger episodes of financial crises. On the other, once a crisis erupts, market participants must have confidence in banks’ reported asset quality metrics in order to regain faith in the financial system. This column shows that accounting standards and prudential frameworks to identify and measure non-performing assets vary widely across countries. This presents a challenge for comparing credit risk across banks and countries, for which the column proposes a range of policy options. 

The timely identification and measurement of non-performing assets (NPAs) remain one of the most consequential responsibilities of banks and prudential supervisors, given their critical role in preserving and, at times, in restoring financial stability. NPA identification refers to the process used to classify an exposure as ‘non-performing’. NPA measurement, in contrast, represents the level of provisions for credit losses required to write down the exposure to its estimated recoverable amount. These assessments drive the credit risk profile of banks, which can materially affect their reported earnings and solvency measures. These earnings and solvency metrics are widely used by market participants to ascertain an institution’s financial health. 

In a recent paper, we take stock of the NPA identification and measurement practices in the US, the EU countries that are part of the Single Supervisory Mechanism, and selected jurisdictions in Asia and the Latin America and Caribbean regions (Baudino et al. 2018). Our goal is to ascertain which role prudential regulation and supervision can play in facilitating the prompt identification and measurement of NPAs. Our study reveals considerable differences across jurisdictions in accounting standards as well as divergent prudential frameworks that govern NPA identification and measurement. 

These differences can materially impact whether or when an exposure gets placed on NPA status. They also affect the amount of provisions needed to write down the carrying value of an NPA. The implications of these divergent approaches are clear: it is difficult for market participants to compare and draw meaningful conclusions from key asset quality and regulatory capital metrics across banks and countries, unless these differences are well-understood. 

Finding #1: Accounting frameworks that underpin the process of identifying and measuring non-performing assets vary across jurisdictions

Published financial statements of banks are typically prepared according to applicable accounting standards. Accounting rules provide a framework to determine the credit quality of an exposure and to set provisions for potential losses. However, there is no single set of globally harmonised accounting rules that guide the NPA identification and measurement process. Multiple accounting standards are used for this purpose, as we explain below.

International Accounting Standard (IAS) 39: International Accounting Standard 39 is referred to as an ‘incurred loss’ model because a loss event must have occurred at the reporting date in order to trigger loan loss provisions. This standard, widely used in the run-up to the financial crisis, has been superseded by an expected credit loss provisioning model. Nevertheless, a number of jurisdictions remain under this standard or an equivalent.

International Financial Reporting Standard (IFRS) 9: On 1 January 2018, the International Financial Reporting Standard 9 became effective and replaced IA39. The new standards require entities to calculate provisions based on expected rather than incurred losses. This lowers the threshold to prompt loan loss provisions. Most jurisdictions in Europe and several other countries have adopted this standard. 

US generally accepted accounting principles (US GAAP): Other jurisdictions, including the US, operate under national standards. Until 2020, US entities will follow an incurred loss approach to setting provisions, which is similar to IA39. Starting in 2020, certain US entities will migrate to the current expected credit loss (CECL) model. This is comparable to the IFRS9, though there are some differences. 

Table 1 compares provisioning requirements among these three models.  

Table 1 Comparing provisions under IAS 39, IFRS 9 and US CECL

Sources: Barclays, IASB, FSI analysis

These various accounting standards can result in differences in the volume and timing of credit loss provisions, which are reflected in bank earnings. When the US migrates to the current expected credit loss model, disparities will persist as this model requires lifetime expected credit losses to be recognised on day 1, while IFRS9 only impose lifetime losses once there has been a significant increase in credit risk. 

Finding #2: Prudential frameworks for non-performing asset identification vary across jurisdictions

The differences in applicable accounting frameworks are amplified by variations in regulatory NPA identification frameworks across surveyed jurisdictions. 

While the Basel Committee on Banking Supervision has recently issued supervisory guidelines on the prudential treatment of problem assets (Basel Committee on Banking Supervision 2017), a number of the jurisdictions we surveyed has not adopted a formal NPA definition for regulatory purposes. In the absence of an explicit definition, several authorities use existing regulatory asset classification frameworks as proxies for NPAs. These frameworks prescribe a combination of quantitative concepts, such as payments that are ‘past due’, and a qualitative, ‘unlikely to pay’, criterion to identify NPA exposures. 

The extent to which supervisors rely on the more forward-looking unlikely-to-pay criterion as opposed to the past-due criterion varies across jurisdictions. This results in differences in whether or when an exposure gets placed in the NPA category. Other variations across jurisdictions stem from differences in the scope of application of regulatory NPA identification regimes, which may preclude certain asset classes from the NPA designation; disparities in the role of collateral in the NPA identification process (some authorities consider collateral, while others assess credit quality independent of collateral); and divergent approaches in the criterion used to exit the NPA category. 

Finding #3: Prudential frameworks on non-performing asset measurement vary across jurisdictions 

Differences in prudential requirements also affect NPA measurement via their impact on provisions. Since NPAs are by definition in, or near default, provisioning outcomes are heavily influenced by the value assigned to collateral, if any. 

In setting requirements for loss provisions, some authorities defer to accounting rules, some impose prudential rules, and others use a combination of the two. Typically, prudential authorities can only impose regulatory rules in relation to provisioning requirements if they have powers to set or override accounting requirements for banks. Figure 1 below shows how the different approaches are implemented across the surveyed jurisdictions. 

Figure 1 Approach used to estimate and recognise loan loss provisions in the profit and loss statement

Note: LAC are Latin America and Caribbean countries, and EU-SSM are countries that are part of the Single Supervisory Mechanism.

Countries in Figure 1 that apply international accounting standards in imposing provisioning requirements generally follow a similar approach. However, both IAS39 and IFRS9 are principles-based standards that demand the use of judgement, particularly with respect to the valuation of collateral that support NPAs. These valuations are heavily dependent on assumptions and require banks and supervisors to estimate the net present value of collateral, by considering the time required to seize and sell collateral on problem loans. These assessments are critical in jurisdictions that have underdeveloped legal-foreclosure frameworks and can result in vastly different provisioning outcomes. 

In contrast, jurisdictions that prescribe regulatory provisioning rules implement a range of approaches that set both the minimum amount of required provisions based on the perceived severity of the problem asset, as well as the methods used to value collateral. Collateral valuation methods set by regulators impose prescribed haircuts to appraised collateral values that can vary based on collateral type, but the required haircuts differ across jurisdictions. 

Finally, the remaining jurisdictions that combine accounting and regulatory provisioning approaches typically require the higher of the two to be recognised in the profit and loss statement of banks. This diverse range of approaches used to determine the size of credit loss provisions can have a material bearing on the reported earnings and regulatory capital levels of banks. 

Policy implications

The varied NPA regimes suggest that simply comparing the reported volume of NPAs and provisioning coverage levels across banks and jurisdictions may be misleading. Recent guidelines by the Basel Committee on Banking Supervision on the first-ever global definition of a non-performing exposure can help to harmonise NPA identification frameworks. 

Beyond this, our study identifies a number of suggestions that may enhance NPA identification regimes. For example, the scope of regulatory NPA identification frameworks should encompass all asset classes and exposures. Greater emphasis in supervision should also be placed on forward-looking factors to determine whether large, wholesale exposures belong in the non-performing category.  

While there is no international standard on NPA measurement, our paper outlines a range of policy options for authorities. These considerations include powers to impose ‘prudential backstops’, i.e. a minimum loss coverage for the amount of money banks need to set aside to cover losses from non-performing exposures, to address situations where accounting provisions on NPAs are deemed inadequate from a regulatory perspective (also Restoy and Zamil 2017). Authorities could also place heightened focus on the prudent valuation of collateral, given the links between estimated collateral values on NPAs and the size of credit loss provisions. 

For market participants, it will remain a challenge to reconcile cross-jurisdictional differences in NPA accounting and prudential regimes, and to understand their implications for the credit risk profile of banks. Prudential authorities cannot eliminate all differences. However, they can foster greater confidence in banks’ reported asset quality, earnings, and regulatory capital metrics by adopting regulatory and supervisory frameworks that support the prompt identification and prudent measurement of NPAs. 

Authors’ note: The views expressed in this column are those of the authors and do not necessarily represent the views of the Bank for International Settlements or the Basel-based standard setters.


Baudino, P, J Orlandi and R Zamil (2018): ‘The identification and measurement of non-performing assets: a cross-country comparison’, FSI Insights on Policy Implementation No 7, Bank for International Settlements.

Barclays Equity Research (2015), “Re-visioning provisioning”.

Basel Committee on Banking Supervision (2017), “Prudential treatment of problem assets – definition of non-performing exposures and forbearance”, April.

Financial Accounting Standards Board (2016), “Financial instruments – credit losses”, Financial Accounting Series, no 2016-13, June.

International Accounting Standards Board (2014), IFRS 9 financial instruments, July.

Restoy, F and R Zamil (2017), “Prudential policy considerations under expected loss provisioning: Lessons from Asia”,  FSI Insights on policy implementation, No. 5, Bank for International Settlements.

1,890 Reads