Banks’ Credit Losses and Provisioning over the Business Cycle: Implications for IFRS 9
The global financial crisis has increased the interest of regulators in the mechanisms reinforcing the inherent procyclicality of banks’ behavior. It has become evident that attention has to be paid not only to the quality of credit exposures, but also to the adequacy of provisioning over the cycle. Provisioning is of crucial importance to the resilience of the banking sector. It serves as a buffer against expected credit losses and significantly influences banks’ profitability, which, in turn, may have an impact on their capital adequacy and lending capacity. Consequently, the question has arisen of how much the regulatory and accounting framework itself contributes to the procyclicality. In fact, numerous studies have found that the accounting framework effective before 2018 (International Accounting Standard IAS 39) was highly procyclical. The new IFRS 9 (International Financial Reporting Standard), which came into force on 1 January 2018, was implemented as a response to this criticism. However, some recent studies indicate that under certain assumptions, provisioning under IFRS 9 may remain procyclical or even exaggerate the procyclicality relative to IAS 39.
In our paper Banks´Credit Losses and Provisioning over the Business Cycle: Implications for IFRS 9, we focus on the procyclicality of banks’ credit losses and provisions in the Czech Republic before 2018 and then discuss the implications of our findings for provisioning for exposures in stage 3 under IFRS 9.1 We consider to be procyclical such credit losses and provisions that tend to decrease when the real economy is growing faster than its potential growth level and increase when it is growing more slowly than its potential growth level or falling. Accordingly, we estimate the sustainable level of lifetime expected credit losses and provisions and examine banks’ procyclicality with an emphasis on potential asymmetries.
The sustainable level in our concept is the level to which credit losses and provisions are supposed to revert in the long term. Credit losses above or below this level should be understood as over- or undervalued, and provisions above or below it should be viewed as excessive or insufficient. Insufficient provisioning may justify the implementation of stricter prudential policies, for example, a higher countercyclical capital buffer rate or additional Pillar 2 capital requirements (in the case of idiosyncrasies between banks). Credit losses that are not covered by provisions will be covered by imposed capital add-ons. Similarly, excessive provisioning may signal the need to implement less strict prudential policies, i.e., to lower the existing countercyclical capital buffer or reduce Pillar 2 add-ons.
We identified significant asymmetries in banks’ provisioning over the cycle (see the table below). In particular, banks seem to recognize credit losses and create provisions with a delay with respect to worsening economic conditions: the increase in credit losses and provisions is concentrated mostly in the later stage of an economic contraction, while we would have expected it to occur in the early stage. Such asymmetry, if it persists under IFRS 9, may have negative consequences and potentially reinforce the inherent procyclicality of banks’ provisioning. Banks are generally less profitable in “bad times”. Postponing the recognition of credit losses and provisioning toward the later stages of a recession intensifies the pressures on profitability and, consequently, may be reflected in banks’ capital and lending capacity. A slowdown in credit growth would feed to the real economy and back to the banking sector, potentially deepening and prolonging the recession.
Another factor potentially exaggerating provisioning procyclicality is a stronger reaction in higher quantiles, indicating that banks with the highest credit risk are the most sensitive to changes in the business cycle. This reaction is apparent in both the upturn and the downturn of the business cycle and may therefore increase the overall amplitude of business cycle fluctuations.
Phases of business cycle | Credit losses | Provisioning |
---|---|---|
later expansionary phase |
moderate effect (-0.217) |
no significant effect |
early contractionary phase (positive and falling output gap) |
no significant effect |
no significant effect |
later contractionary phase (negative and falling output gap) |
strong effect (-0.586) |
strong effect (-0.286) |
early expansionary phase (negative and rising output gap) |
strong effect (-0.667) |
strong effect (-0.357) |
Note: The numbers in parentheses are estimated coefficients for the given relationships. The assessment of the strength of the effect in individual phases of the business cycle is relative to the average effect over the entire business cycle, which is -0.391 for credit losses and -0.160 for provisioning. For full estimation results, see Table 2 in the paper Banks´Credit Losses and Provisioning over the Business Cycle: Implications for IFRS 9.
The delayed transfer of exposures between stages and the pronounced impact in higher quantiles may significantly contribute to a sharp increase in lifetime expected credit losses and provisions in response to an unexpected deterioration in economic conditions. Expected credit loss models, which replaced incurred loss models under IAS 39, rely heavily on forward-looking information about future macroeconomic developments. This forward-looking information is produced by models which often underestimate the probability and severity of future economic downturns.
Macroeconomic forecasting models are usually able to predict some degree of mild economic slowdown, but not a severe deterioration. Macroeconomic projections are usually revised only after the economic downturn has already occurred. This may trigger an abrupt transfer of exposures between stages and lead to a “cliff effect”2 of potentially larger magnitude relative to IAS 39. The actual magnitude of this cliff effect would depend largely on how banks implement the new standard and its individual parts. It might take some time, or even be impossible, for banks to come up with an adequate modeling approach appropriately incorporating inherently inaccurate macroeconomic projections and at the same time accurately estimating expected credit losses. It is therefore likely that the delay in the recognition of credit losses and provisioning under IFRS 9 will persist in the near future, leading to exacerbated procyclicality.
1 Under IFRS 9, credit exposures are divided into three stages. At the inception of the loan, the exposure is immediately categorized as stage 1 and an impairment allowance is set to cover losses at the 12-month horizon. Once a significant increase in credit risk occurs, the exposure is transferred to stage 2 and a credit impairment allowance is set to cover the credit losses that are expected to materialize over the lifetime of the asset. The transfer to stage 3 is triggered by the occurrence of a loss event. At this stage, credit impairment allowances should still cover lifetime expected credit losses. Stage 3 is conceptually the most similar to the incurred loss approach under IAS 39, which already required estimation of lifetime expected credit losses for impaired loans.
It is important to note that banks have only been applying IFRS 9 since the beginning of 2018, which limits the assessment of its potential effects. A full evaluation will be possible once banks gain experience in provisioning according to IFRS 9 and data become more available and reliable. In our research paper, we perform our analysis on the sample prior to the implementation of IFRS 9; our results are therefore indicative mostly of the exposures in stage 3, as explained in more detail in the paper.
2 A “cliff effect” can be understood as an abrupt transfer of exposures between stages in response to the arrival of unexpected information indicating a weakening of the economic conditions.