Does the Disclosure of Proprietary Information About Loan Losses Pose a Threat to Financial Stability?

If banks are slow or reluctant to publish negative information about loan losses, should the financial supervisory authorities [FSA] then ensure that the information is revealed to the financial markets earlier by publishing private information obtained from ordinary bank inspections? The release of such information has the potential to increase efficiency (Goldstein and Sapra 2014) but could also cause instability if there are large contagion effects. The question relates to a larger debate on the relationship between transparency and stability including how much information bank regulators should disclose about individual banks (Morris and Shin 2002; Morrison and White 2013; Goldstein and Sapra 2014).

Most developed countries, including the US (Deyoung et al. 2001; Bushman 2014) do not release information from ordinary on-site inspections, and the information they do release is in the form of formalized reports, which in most cases cover several banks. The purpose of these reports is to analyze banks from a prudential and systemic view, e.g. if they have sufficient capital, and to assess the robustness of the financial system, and not to provide timely information to financial markets. The asset quality review (AQR) by the European Central Bank in 2014 (European Central Bank 2014) and subsequent reports is an example of this type of report. Prior literature has documented that such one-time supervisory disclosures have capital market consequences (e.g., Bischof and Daske 2013; Durrani et al. 2022). In the US, bank supervisors are required by law to make a public announcement of formal enforcement actions taken against specific banks (Jordan et al. 2000; Kleymenova and Tomy 2022). However, since formal enforcement actions are only issued against the most troubled banks when the US bank supervisors believe that immediate remedial action is needed to avoid failure, even in the US – an advocate for bank transparency—little is known about market reactions and spillover effects from the release of information from ordinary on-site inspections.

In this paper we focus on the Danish FSA’s public disclosure of ordinary inspection reports that include required increases in loan loss reserves—one of the key variables in assessing the performance and risk of banks. The FSA’s announcement that a particular bank’s loan loss reserves were found to be insufficient, and that the supervisors require corrective measures by the bank is an enforcement action, and prior literature has documented negative market reactions and spillover effects to supervisory disclosures of enforcement actions (e.g., Jordan et al. 2000). However, we use reports from the FSA that are part of an established, standard, and recurring inspection system where the sanctions in our case are directly related to the correction of insufficient loan loss reserves and do not normally involve more than this.

We test if announcements of additional loan loss reserves by the FSA after on-site inspections contain new information to the financial markets during our sample period 2009–2020. We find clear and strong evidence of average negative returns in the inspected banks when the FSA announces that it requires increases to the loan loss reserves. The results thus indicate that the announcements contain new negative information for the financial markets.

The announcements of additional loan losses by the FSA may lead to contagion effects in peer (i.e., non-announcing) banks with implications for the financial stability of the entire banking sector. We analyze potential contagion by first analyzing the effects on all banks as well as for high/low risk banks and large/small banks. Except for large banks, we find statistically significant negative contagion effects. Then we analyze if the contagion effects depend on similarities or differences between the announcing banks and the peer banks. We specifically look at contagion effects in high and low risk banks from FSA announcements for other high and low risk banks, and at contagion effects in large and small banks from FSA announcements for other large and small banks. We find contagion effects in high-risk banks triggered by FSA announcements concerning other high-risk banks. However, there are no significant effects on low-risk banks stemming from FSA announcements related to either high-risk or low-risk banks. In the case of small banks, we identify contagion effects from FSA announcements concerning other small banks, yet no effect is observed when the announcement pertains to a large bank. Conversely, for large banks, we find no contagion effects irrespective of whether the FSA announcement pertains to a large or small bank. Although we do find significant contagion effects, the economic magnitude of the contagion effect is small. The largest effect is for high risk banks, but the FSA announcement of additional loan losses regarding another high risk bank of e.g. 100%, i.e., certain bankruptcy, only has a contagion effect on other high risk banks of around 4%.

In general, contagion effects may stem from economic fundamentals and behavioral factors. For economic fundamentals, equivalent banks are exposed to the same economic conditions, and hence information about the announcement bank is relevant for its peer banks (e.g., Bischof et al. 2021). This information would reach the financial markets eventually, and the FSA announcement consequently only accelerates the process. Alternatively, from a regulator’s point of view, it is more serious if the contagion is caused by investors who misinterpret the information or overreact to the situation. We are not able to fully distinguish between fundamental and behavioral explanations in our study. However, we provide results that indicate that the Danish capital market reacts rationally to the information, which points toward the fundamental explanation.

Our sample period includes the recent financial crisis, and Flannery et al. (2013) show that the opaqueness of banks increases during crises, which in turn increases the likelihood of contagion effects from the release of information from the supervisory authorities. Thus, we find a small effect in relation to the key area of loan loss reserves during a period in which the effect should be the strongest. Overall, our study disproves the concerns that have been expressed about potential instability of the financial systems if the FSA publicly disclosed the outcomes from ordinary inspections of banks about loan losses.

Our study adds to the scarce literature about individual and contagion effects from announcements of increases in loan loss reserves where the announcements come from an authoritative external source and not the banks themselves. Our paper contributes to the literature on regulatory disclosures and financial markets by leveraging Denmark’s unique environment, where the Danish FSA publicly discloses bank inspection results. We provide new insights into the market reactions and contagion effects of these disclosures, focusing on ordinary inspections rather than one-time supervisory reports or formal enforcement actions. While our FSA announcements share a common focus on specific banks' asset quality with previous research, market reactions may differ. Stock markets might react differently to loan loss provision adjustments in ordinary FSA inspection reports compared to stress-test results. Individual inspection reports offer specific, timely information about a single bank, potentially causing a sharper stock price reaction. In contrast, stress-test results released simultaneously for multiple banks (see for example the EBA(2023) stress-test) can potentially dilute the immediate impact on the specific bank's stock price but increase contagion effects because investors may focus on the collective results and systemic implications rather than individual banks. Additionally, individual inspections can potentially surprise the market more than pre-announced stress-tests where the methodologies, scenarios and key assumptions usually are known. Ordinary inspections usually indicate less severe issues than formal enforcement actions, and the market may hence view these as issues that can be addressed without major disruptions, leading to a more moderate impact on stock prices. We show a link between share price reactions and the size of insufficient loan loss reserves, suggesting that the reaction is related to proprietary information about loan losses and not just because investors perceive them as enforcement actions as such. Overall, our study contributes by documenting the immediate stock market consequences of supervisory disclosures of proprietary information about loan losses in a system with open ordinary inspection reports.

The remainder of the paper is structured as follows: Section 2 includes a review of related research and our hypotheses. Section 3 explains our research design. Section 4 provides a brief description of the Danish institutional setting. Section 5 describes the data and contains descriptive statistics. Section 6 presents the results of the tests of market reactions to the FSA announcements by investors in the inspected banks. Section 7 presents the results of the tests of contagion effects, i.e., the tests of market reactions to the FSA announcements by investors in non-announcing banks. The last section summarizes the main conclusions and explains certain limitations of the study.

Loan loss reserves provide a cushion against future losses on the loan portfolio. In general, banks review their loan portfolios on a regular basis. If there is any objective evidence suggesting that a financial asset or group of financial assets is impaired due to, for instance, significant financial difficulty (increase in default probability and/or decrease in expected recovery rates) of an obligor or risk of breach of contract, such as default or delinquency in interest or principal payments, banks will, in accordance with the international accounting standard IAS 39, recognize a loan loss provision in the income statement with a corresponding increase in loan loss reserves in the balance sheet. The current international accounting standard IFRS 9 even requires the provision to be recognized at the time of initial recognition of the loan. Additional loan loss provisions decrease the book value of equity of the bank and its solvency ratios and may, if large enough, close the bank.

Previous research on market reactions to changes in loan loss reserves has focused on announcements of changes to loan losses outside the financial reports and mostly on losses associated with the international debt crisis commencing in 1982. Findings regarding the reaction of the financial markets to announced increases in loan losses are mixed. Grammatikos and Saunders (1990), Musumeci and Sinkey (1990) and Wahlen (1994) find a positive effect, Liu and Ryan (1995) find a positive effect for banks with sizable frequently negotiated loans, while Cornell and Shapiro (1986), Bruner and Simms (1987), Smirlock and Kaufold (1987), Mansur et al. (1990), Lancaster et al. (1993), and Docking et al. (1997) find a negative effect from loan loss announcements. 

Deyoung et al. (2001) study whether supervisory authorities’ examinations of US banks produce useful information that is not already reflected in market prices (spread) for subordinated debt and find that the private information in the reports predicts future spreads. Thus, on-site inspections produce value-relevant information about the future safety and soundness of banks several quarters before this information is reflected by the markets in debenture prices.

Bischof et al. (2021) study the more recent financial crisis (2007–2009) and find that banks were late and reluctant in reporting the additional loan losses. They also find that the financial markets’ reaction to banks’ own disclosures about subprime exposures is negligible and statistically insignificant. The insignificant reaction either suggests that the banks did not disclose relevant material information or that investors and analysts formed expectations based on other sources about the banks’ troubled assets ahead of their disclosures.

The present paper differs from most of the existing literature in one important way, which is that the announcements are issued by the supervisory authority after an inspection of a bank, and they are therefore both public and independent of bank managers' own disclosure decisions. The paper by Jordan et al. (2000) is an exception that also uses announcements from bank supervisory authorities about specific banks. They analyze stock market reactions to announcements of formal supervisory actions against large US banks and find that investors react negatively to the announcements and that they can cause spillover effects. However, formal supervisory actions are only issued when the US bank supervisors discover serious “life threatening” problems related to unsafe and unsound practices in the bank which require immediate remedial action (Jordan et al. 2000). Our setting is different. We use reports from the Danish FSA which are part of an established, standard, and recurring inspection system where the sanctions in our case are directly related to the correction of insufficient loan loss reserves and do not normally involve more than this. The Danish setting has previously been studied by Kjeldsen and Raaballe (2015). However, in contrast to our study they include all “bad news” announcements which in addition to required increases in loan loss reserves include announcements of negative changes in a bank’s solvency unrelated to FSA reviewed loans and all announcements of low or insufficient solvency coverage, and their study covers a shorter time period than ours. Kjeldsen and Raaballe (2015) find a negative share price effect from this broad collection of “bad news announcements”. In our study we select the inspection reports in which the FSA based on a review of a selection of the bank’s loans during the inspection announces that the inspected bank needs to increase its loan loss reserves, and we analyze if the announcements contain value-relevant news to the investors. Given the varied findings in prior studies, encompassing both positive and negative market reactions to disclosures about changes in loan loss reserves, our first null hypothesis is formulated as:

Goldstein and Sapra (2014) discuss positive and negative effects of public disclosure of supervisory authorities’ individual bank stress tests. More openness implies a broader information base and hence more informative share prices. Thus, an important positive effect is the potential to promote bank stability through more informed decision-making. However, there could also be negative effects associated with public disclosure of information concerning a bank’s financial condition. Financial market participants are typically not only interested in the actual capital or solvency of a bank (i.e., the bank’s fundamental value); they also have strategic concerns and care about the beliefs of other market participants as to the bank’s financial condition (Goldstein & Sapra 2014). Morris and Shin (2002) show analytically that in such environments, public disclosure of information gathered by regulators, for example, leads to underweighting of the market participants’ own private information and overreaction to the public information. In other words, there is a risk of negative information transfer from one investor to the other – perhaps even unrelated to bank fundamentals. Bank regulators are particularly sensitive to correlations and negative information transfer from disclosures due to their shift in emphasis from a micro-prudential approach to a macro-prudential perspective where the focus is on limiting financial system-wide distress (Beatty and Liao 2014).

Several channels for contagion exist: systematic or macro-economic effects that are relevant to other banks as well, information channel effects which refer to the transmission of negative information or sentiments through various channels in financial markets (e.g., concerns about the content and credibility of other banks’ financial statements) and counterparty risk. While these channels all imply negative returns for other banks from an announcement of increased loan loss reserves in one bank, there is also a competitive channel with positive effects to the peer banks (Egginton et al. 2010).

Most of the existing research analyzing contagion effects has focused on the banks’ own release of information, e.g. loan losses, or has used the closure of banks, including the closure of Lehmann Brothers, as the event (Aharony and Swary 1983, 1996; Docking et al. 1997; Akhigbe and Madura 2001; Jorion and Zhang 2007, 2009; Egginton et al. 2010; Aragon and Strahan 2012; Chakrabarty and Zhang 2012; Fernando et al. 2012; Helwege and Zhang 2016; Bischof et al. 2021). Most of the existing research on contagion is consequently from financial distress settings in which counterparty contagion is an important factor. Our setting is different. We focus on contagion from announcements of loan losses that are based on the supervisory authorities’ ordinary on-site inspections. The announcements do not depend on the banks’ own disclosure decisions. In this setting, the information effect is probably significantly stronger than the counterparty effect since only the probability of credit loss is affected, whereas in the existing studies, actual losses are involved. The setting is, in effect, a restatement setting, and Xu et al. (2006) and Gleason et al. (2008) find that there are indeed contagion effects associated with restatements. Moreover, we do not see any pattern in our sample of FSA announcements on increases in loan loss reserves that indicate that contagion effects systematically relate to a specific sector, as the announcements typically just state the amount of the additional loan loss provision and do not specify the underlying sector to which the additional provision relates.

Since the different channels involve both positive and negative effects, the contagion (null) hypothesis we are testing is given by:

Our study of individual and contagion effects concentrates exclusively on immediate stock market effects. Such immediate adjustments to the stock price capture market sentiment, i.e., how investors feel about the information released based on the updated financial information, such as changes in asset quality or expected future earnings. However, these stock market returns likely fail to capture more unpredictable long-term effects such as depositors potentially withdrawing funds, or operational impacts on the peer banks including any strategic adjustments they might make in response to the announcement. Moreover, by concentrating on immediate stock market effects we do not consider other consequences of supervisory disclosures that share prices fail to capture. We discuss such issues in more detail in the concluding section.

Hypothesis 1 is tested using a standard event study methodology (MacKinlay 1997; Bartholdy et al. 2007). The standard event study methodology uses the market model to calculate the abnormal return (AR_i,t) for each bank i on day t. The market model parameters (α_i; β_i) are estimated using the benchmark index (r_mt) for Copenhagen Stock Exchange, (OMXCB). OMXCB is a well sector diversified free float-capitalization index designed to act as a transparent and liquid benchmark (The NASDAQ OMX Group 2009). Like for example Wang (2014) we use an estimation period of 200 trading days and we use an initial event window of 21 days, which is subsequently reduced to a shorter, more focused window. Our tests of abnormal returns are based on AAR and CAAR. AAR is the average abnormal return across all banks from our sample on day t, and CAAR is the aggregated AARs over the interval t [0; 1] and t [0; 2] with day t 0 being the FSA announcement day. Stock prices, market value and trading volume for each bank are retrieved from Datastream. The shares on the Copenhagen Stock Exchange suffer from thin trading. As a counteracting measure to the thin trading issue, we utilize trade-to-trade returns (Maynes and Rumsey 1993; Bartholdy and Riding 1994; Bartholdy et al. 2007).

Next, we classify each announcement as one of three types based on the information in the FSA announcement:

We distinguish between the three different types of announcements and analyze whether the reactions in the event window are related to the severity of the announcement and to the size and riskiness of the announcing banks. We estimate the following model in addition to a baseline model without the risk and size variables:

CAR_it is the cumulative abnormal returns calculated as the sum of bank i’s abnormal returns over the interval t[0; 2]. For Type 1 and Type 2 announcements, we calculate two measures to analyze the strength or severity of the announcement: Equity-loss-ratio and Mk_Equity-loss-ratio. The first measure scales the announced additional loan losses with the book value of equity at the beginning of the fiscal year, and the second measure uses the market value of equity 11 days before the announcement as denominator. The amounts of additional loan loss reserves were collected from the announcements, and we obtain balance sheet and income statement data for the period 2004–2019 from the FSA homepage.¹ Ln(Size) is the log of total assets in the bank. Risk is measured by the Z-score (Lepetit and Strobel 2015; Berger et al. 2014; Beck et al. 2013; Delis et al. 2014):

Higher levels of equity to total assets, higher levels of Return on Assets (ROA), and lower levels of standard deviation of ROA reduce the overall level of risk of the bank and increase the Z-score, thus a high Z-score implies less risk. To ensure that the information is available to financial markets at the time of the announcement, the various performance measures in the paper are calculated using data from the previous fiscal year, e.g., ROA for an announcement in 2015 uses accounting data for 2014. The standard deviation of ROA is calculated in a rolling window over the previous five years. However, for banks with less than five years of data, we require a minimum of three years to be included. Thus, the Z-score for an announcement in 2012, for instance, is calculated using data for 2007 to 2011. In line with common practice in existing literature we also log transform the Z-score.

The basic contagion arises if the regulatory authority issues an announcement of a demand for increased loan loss provisions in bank A and this affects the share prices of banks B,C,…. Thus the contagion effect is measured using the returns of banks B,C,… following an announcement relating to bank A. Since the regulatory authority also issues announcements for private banks, we can include these in our sample, i.e., the announcing bank A does not have to be listed, only banks B,C,….. The general approach for the contagion tests involves forming an equally weighted portfolio spanning January 1, 2009, to December 31, 2020, comprising all listed banks on a given date. For each date, a bank is included in the portfolio if it has no announcement at that time. If the FSA does make an announcement for a bank, it is excluded from the portfolio for the announcement date and the following two days but included otherwise. The portfolio return at a given time consequently reflects the returns of banks without announcements at that time. The contagion analysis incorporates the Equity_loss-ratio for Type 1 and Type 2 announcements and a dummy variable for announcements without specified amounts.

The basic contagion model (Eq. 2) is as follows:where Type1_ Equity-Loss-ratio⁰_it at time t represents the Equity-loss-ratio value for announcing bank i at the announcement date and is zero if no bank in the sample has an announcement at time t. For days with multiple announcements, the values of Equity-loss-ratios are summed. The total effect of announcements is captured by the sums of coefficients for Type 1 (γ₀ + γ₁ + γ₂) and for Type 2 announcements \(\left({{\varvec{\lambda}}}_{0}+{{\varvec{\lambda}}}_{1}+{{\varvec{\lambda}}}_{2}\right)\), analogous to traditional Cumulative Abnormal Returns (CAR) analysis in event studies.

We augment the basic contagion model in Eq. (2) and differentiate between FSA announcements for listed and private banks, considering the generally smaller size and lesser attention in the financial press given to private banks, which may result in reduced contagion effects. To address this, we split the Type_1_equity_loss-ratio into two variables representing FSA announcements for listed banks (Type1_Lis_Equity-Loss-ratio) and private banks (Type1_Priv_Equity-Loss-ratio). Due to insufficient observations, we are unable to split the Type_2_Equity_Loss variable into listed and private announcements. This leads to the revised Model (3):

In our final contagion analyses we augment the basic contagion model in Eq. (2) with information about the size and riskiness of the announcing banks. That is the independent variables are respectively Type1_Equity_Loss-ratio_Large and small and Type1_Equity_Loss-ratio_High and Low. This enables us to test for contagion effects in peer banks that share the same characteristics.

Danish financial institutions, both listed and non-listed banks and savings banks, are supervised by The Danish Financial Supervisory Authority [FSA]. This right and duty is conferred by law (Sect. 346 of the Financial Business Act). Denmark is an EU member, but it does not have the euro as its currency. Consequently, Denmark does not automatically participate in the Banking Union, and so far, Denmark has not applied to the ECB to enter a so-called “close cooperation” concerning the supervision of credit institutions. Hence, Denmark does not participate in the Single Supervisory Mechanism (Ministry of Industry 2019). Despite this, the Danish FSA participates in the work of the European Banking Authority (EBA), including the work related to supervisory methods, and as a general rule the Danish FSA follows all EBA's guidelines and recommendations (Christiansen et al. 2017, p. 581 and p. 614). This includes the guidelines on common procedures and methodologies for the supervisory review and evaluation process (SREP) (European Banking Authority 2018). Consequently, the approach and the procedures used by the Danish supervisory authorities are closely aligned with the approach and the procedures used elsewhere in Europe.

Our study focuses on ordinary inspections, which primarily concentrate on the FSA’s assessment of risks within the lending area, such as systems and procedures for granting and following up on loans. The Danish FSA also conducts "thematic" and "function" inspections, which include evaluations of IT security, money laundering detection systems, and management of interest rate risks. These elements are typically not part of ordinary inspections. During an ordinary inspection the FSA reviews a selection of the banks' loans and assesses the sufficiency of their loan loss reserves. We select inspection reports where the FSA specifically concludes that these reserves are insufficient.

Ordinary inspections may uncover weaknesses in credit control, compliance, internal audit, and documentation, etc. often tied to loan loss reserves. While the primary focus is on loan loss reserves, these reports for instance also assess the bank's solvency requirements. The FSA may critique a bank's solvency ratio methodology or its adequacy, regardless of loan review findings. Thus, market reactions might not be entirely due to the loan loss reserves section in the inspection reports. However, our paper shows a link between share price reactions and the magnitude of insufficient loan loss reserves, suggesting that this prominent information significantly influences the market response.

Once a year the FSA plans next year’s inspections based on general guidelines from the board (Christiansen et al. 2017, p. 588). The FSA inspects banks depending on their size and perceived risk. Small institutions with an average risk profile are inspected about once every four years, whereas the large institutions are inspected several times every year (Kjeldsen and Raaballe 2015). Banks subject to so-called "enhanced supervision” are generally inspected every year. These are banks for which the Danish FSA, based on received reports, previous inspections, or other information, assesses that special attention is needed. Kjeldsen and Raaballe (2015) describe the practice in the inspection process in the following way: One or two months prior to the inspection, the FSA contacts the bank and requests the material required for the inspection. During the inspection this material is analyzed by the FSA and the management of the bank. A couple of weeks after the inspection, the FSA meets with the management of the bank, and the FSA presents its conclusions. The FSA prepares an extensive report as well as a short summary. Since 2010, it has been mandatory for both inspected Danish banks and the supervisory authority to publish the summary on their homepage. Prior to 2010, listed banks were required to publish all information relevant for the pricing of the shares. In this paper, we assume that increases to loan losses are indeed considered relevant for the pricing of shares, and the listed banks are therefore required to publish the information immediately after receiving the final report from the FSA.

Denmark introduced the International Financial Reporting Standards (IFRS) in 2005, which changed the accounting rules applicable to loan losses. Before 2005, the amount recognized as loan impairments was based on management’s estimates of future losses, and those estimates were to some extent applied as a means to smooth earnings by building up reserves in years of prosperity to be used in periods with downturns in the economy.² The implementation of the International Accounting Standards in 2005 introduced an "incurred loss model", where loans are impaired and impairment losses are consequently recognized if, and only if, there is objective evidence that a “loss event” has occurred after the initial recognition (IAS 39.59) (IASB 2013). In the first years after the adoption (2005–2007), the situation in Denmark was unusual with net negative loan loss provisions (i.e., positive net income effects) for the whole sector due to the reversal of the too conservative loan loss reserves built up by the banks previously (Danmarks Nationalbank and Abildgren 2010, p. 155, Novotny-Farkas 2016). Effective from 1st of Jan. 2018, the incurred credit loss model was replaced by an expected credit loss model in accordance with the regulation in the international financial reporting standard IFRS 9 (IASB 2014).³ This change leads to earlier loan loss recognition and allows for more managerial discretion.

Close to the beginning of our effective sample period in 2010, Denmark had a total of 123 banks and savings banks. The supervisory authority divides them into four groups based on size. We exclude 27 very small banks and savings banks with working capital below DKK 250 m (approximately EUR 33 m.) and focus on the remaining 96 financial institutions in the three largest groups. Size group 1 is defined as banks with “working capital” (deposits, issued bonds, subordinated debt, and equity capital) above DKK 50 bn, group 2 has working capital between DKK 10 and 50 bn, and size group 3 has working capital between DKK 250 m and DKK 10 bn. In 2010 the six largest banks account for about 86% of the total assets and about 80% of the deposits and issued loans. 12 banks belong to the second largest group with about 7% of the assets and about 9% of the deposits and loans. The 78 smallest banks included in our analysis account for roughly the same as the second largest group of banks.

Data from the World Bank, Financial Structure Database⁴ (non-reported) shows that in terms of market concentration, provisions to non-performing loans, ratio of defaulting loans to total gross loans and regulatory capital to risk-weighted assets the Danish banking sector is comparable to other countries in the Euro area in the period leading up to the beginning of our sample period.

Since 2010, the FSA has been required to publish their findings on their homepage.⁵ Again, the dates of the publication on the homepage were cross-checked with OMX NewsClient and the earliest date was used.

Table 1 reports the number of FSA required announcements of increases in loan loss reserves for listed and non-listed banks and savings banks operating in Denmark from 2009 to 2020.⁶ During this period, we found a total of 88 announcements for listed banks. Four of these announcements pertain to bankrupt banks where the bankruptcy and halt in trading occur on the same day as the loan loss announcement. These four announcements are therefore excluded from the event study (Hypothesis 1) but included in the contagion study (Hypothesis 2). Thus, the event study includes 84 announcements and the contagion study 88 announcements for listed banks. 76 of the 84 dates were single announcement dates, while on four dates there were two announcements.

FSA announcements for non-listed financial institutions are only available from 2010 when the FSA was required to publish inspection reports on their homepage. From 2010 to 2020, we found 73 announcements for non-listed banks and savings banks. Non-listed financial institutions are typically smaller than listed institutions, and the lower percentage of announcements for non-listed institutions is probably due to fewer inspections by the FSA. From Table 1 it is clear, that there has been a significant net reduction in the number of financial institutions in Denmark during our sample period: about 50% of all financial institutions that were in the market in 2009 are gone in 2020. This is due to a merger wave and to a few bankruptcies in the aftermath of the financial crisis.

For Hypothesis 1 we use announcements for listed banks, whereas for Hypothesis 2 we use announcements for both listed banks and private (non-listed) institutions.

Table 2 reports the number of FSA announcements for individual listed and non-listed institutions. In the group of banks that faced an announcement during our sample period 2009 to 2020, most received only one (11 listed and 17 private institutions). Nine listed and five private institutions faced four announcements or more. Out of the 43 different banks listed at some point during the period, nine banks faced no announcements by the FSA.

The number of announcements by type and bank size groups are reported in Table 3. Most announcements are of Type 1, i.e., announcements where the amount of additional loan losses is specified, and where the announcement is new information to investors. 128 out of 161 announcements are of Type 1, whereas we only have 20 Type 2 announcements and nine announcements without an amount specified. For the listed banks, eight announcements came without an amount, whereas only one no-amount announcement was for the private banks. One example is an announcement from 30 March 2012 regarding the listed bank Nordea. In the announcement from the FSA, they write: “…In impairment calculations, the bank reduced the impairment amounts with expected cash flows, which in several cases were too large. The Danish Financial Supervisory Authority instructed the bank to ensure that the bank only takes into account realistic expected cash flows in the impairment calculations and only in cases where it is not likely that the customer will have to cease farming in the short or long term due to financial difficulties..” In our analysis these observations, when relevant, are included as separate announcements, but given that we only have nine events of this type, the results are subject to some uncertainty.

For listed banks the average Type 1 Equity-loss-ratio is 12.03%, and the median value is 4.24%, reflecting that we have included four banks declaring bankruptcy with large Equity-loss-ratios. The mean of the Mk_Equity_loss_ratio is 39.42%, and the median value is 8.95%. The higher Equity-loss-ratios based on market values of equity than book values of equity reflect a smaller denominator.

We measure the leverage of each bank by the Equity ratio and calculate two measures. The first measure, Equity-ratio, uses book values of equity divided by total assets, both measured at the end of the previous fiscal year, and the second measure, Market-Equity-ratio, uses the average market value of equity during the month of December of the previous year divided by total assets at the end of the previous fiscal year.

Similar to the Equity-loss-ratios, the Type 1 Equity-ratios based on market values (6.78%) and book values (9.91%) differ significantly for listed banks. Together with the observation that market and book values of equity are similar (around 12%) for listed banks without any announcements during our sample period, this suggests that market values decrease prior to the announcement or that financial markets anticipate the announcement. This is in accordance with findings in Bischof et al. (2021), who analyzed a sample of 10 important US banks during the financial crisis in 2007–2009. They find that credit default swap spreads were often substantially increased before the banks themselves released information about their subprime exposures, funding structures and interest rate sensitivities. This indicates that investors used other information sources to form their expectations about the banks’ losses ahead of their disclosures.

For Type 2 announcements, where the loan loss information is already included in the book values, the Equity-loss-ratios for listed banks are very close based on book and market values—approximately 6% and 7%, respectively—suggesting that financial markets adjusted the market values at the release of the financial statements containing the additional loan losses.

In terms of risk, leverage and performance, listed banks facing Type 1 and Type 2 announcements do not seem to differ significantly (average Z-score of 24 and 27 for type 1 and type 2 respectively, Equity-ratios close to 10%, Market-Equity-ratios close to 7% and average ROA around 0% for both.

For private banks the average Equity-loss-ratio for Type 1 announcements is 12.64%, and the median is 4.09 both of which are very close to the values for listed banks. For Type 2 announcements, the Equity-loss-ratio is approximately 3% for private banks and thus about half the value for listed banks. An average Z-score around 27 for Type 1 announcements and 34 for Type 2 indicates that the private banks have slightly less risk than the listed banks. There appears to be very little difference in risk and performance between the private banks facing Type 1 announcements and the ones not facing any announcements in the sample period, suggesting that many of the FSA inspections leading to announcements are regular inspections.

In summary, for both listed and private banks, Type 1 announcements where the announcement is new information to investors, constitute a significant part of the book value of equity – approximately 12% on average. For Type 2 announcements, where the loan loss information has already been incorporated in previous information published by the banks, the average amounts are smaller, particularly for private banks. Our results suggest that market values decreased prior to the announcement of Type 1 announcements whereas for Type 2 announcements the financial markets appear to have adjusted the market values at the release of the financial statements containing the additional loan losses. Despite of this there may still be announcement and contagion effects associated with the FSA’s release of their inspection reports with information about insufficient loan loss reserves. This will be analyzed in the following sections.

The first hypothesis is given by:

Table 4 provides statistics for the abnormal returns. Unreported results show that most of the impact is centered around the announcement days 0, + 1 and + 2 and throughout the remainder of the paper we will therefore work with an event window of these three days. There is also evidence of skewness and kurtosis, and consequently we also need to use non-parametric statistics when testing Hypothesis 1. We apply a battery of test statistics (T1-T7) based on Maynes and Rumsey (1993) and Bartholdy et al. (2007), which, for instance, take into account event-induced changes in volatility and cross-correlations. We base our conclusions on the overall picture, which in our case is very consistent.

Table 4 reports the results of the tests of abnormal average returns and cumulative average abnormal returns. The table shows clear and in general strong evidence of negative returns over the period 2009–2020. The results in Table 4 consequently reject H₁.

The average abnormal returns are -3.16% on the announcement day and -1.17% and -0.58% on the following two days with a CAAR of -4.90%. CAAR for other windows (not reported in Table 4) are t[-1; 1] -4.33%, t[-2; 2] -5.28%, t[-3; 3] -5.12%, and t[-5; 5] -5.32%. Since our study is about the FSA correcting previously released financial information, a comparison with results from the restatement literature is relevant. Gleason et al. (2008) report that the mean three-day announcement period abnormal stock return (t[-1; 1]) for firms that announce ‘‘securities’’ misstatements, which they associate with financial services firms, is -11.7%. Several papers have examined the market reactions to restatement announcements by non-financial firms in which errors or irregularities in the financial statements are corrected. Early work shows substantial negative market reactions. Palmrose et al. (2004) find -9%, GAO (2002) finds nearly -10%, and Dechow et al. (1996) find -8%, whereas more recent studies find more modest reactions to restatements: GAO (2006) find -2%, Atwood et al. (2010), Files et al. (2014), Myers et al. (2013), and Herly et al. (2020) all find -1–2%.

In sum: The magnitude of the market reaction in our study seems smaller than the reaction found in Gleason et al. (2008), but remarkably higher than the market reactions found in more recent studies. Interestingly, the mean three-day return for t [-1; 1] in this study about announcements of insufficient loan loss reserves after an ordinary on-site inspection by the banking supervisory authorities is close to the -4,97% mean three-day return found by Jordan et al. (2000) in their study about announcements of formal supervisory actions in large US banks during the 1989–94 period. The market reaction depends on the information available to investors before the announcement. Hence, the comparable reaction to the more severe US announcements that indicate a high potential for failure and an immediate need for remedial actions compared to the announcements of insufficient loan loss reserves discovered during ordinary inspections could indicate that market participants in the US to a higher extent were able to form expectations about the banks’ serious financial troubles ahead of their disclosures based on other sources than were Danish market participants about the quality of the banks’ loan loss reserves. Alternatively, the comparable reaction could indicate that investors have become more sensitive to information from the supervisory authorities in our more recent sample period—perhaps because of the financial crisis.

We use Eq. (1) to analyze whether the reactions in the event window are related to the severity of the announcement and to the size and riskiness of the announcing banks. The results are presented in Table 5. Models 1–4 all use CAR estimated over t[0; 2] as the dependent variable and vary the scaling factor for the equity loss ratio: models 1 and 2 scale the required additional loan loss reserves by the book value of equity, whereas models 3 and 4 use the market value of equity 11 days prior to the announcement as the scaling factor.

The coefficient for Type 1 announcements, \({\gamma }_{2}\), is negative and significant for all models, providing clear evidence that the announcement contains new information, and that the reaction is a function of the size of the announced loan losses.

The effects of risk and size of the announcing banks are explored in Models 2 and 4. Regardless if book or market values are used as the scaling factor, Ln(Zscore) is positive and significant, thus the negative reaction to loan loss announcements is larger for risky banks compared with less risky banks. These results are consistent with Deyoung et al. (2001), who study bank inspections by supervisory authorities in the US and find that the market seems to value inspection information for troubled banks more highly than for healthy banks. Neither Ln(Size) nor the dummy variable for announcements without any value is significant and thus the reaction for small and large banks is the same.

Supervisory authorities may be more concerned about releasing information from inspection reports if the investors seem to overreact to information. It is not possible to test for overreaction directly, but we can set up some wide boundaries to indicate whether or not an overreaction is taking place. Here, we first use the fact that an additional loan loss that (net of tax) reduces equity by 1% should also reduce the price of the shares by 1%. In other words, 1 Danish krone (DKK) of additional (unexpected) loan losses before tax reduces the market value of equity by (1-tax rate), given that the loan losses are tax-deductible. An overreaction consequently requires a coefficient larger than approximately 0.75 in our setting.⁷ The coefficients for Type 1 announcements in model 3 and 4 which both use Equity loss ratios based on market values shortly before the release of the FSA announcements are well below and significantly different from 0.75, thus showing no evidence of overreaction. To complement the observations related to Type 1 announcements, Type 2 announcements can be used to obtain insights into whether the market seems to react rationally to the information. Type 2 announcements contain no news regarding loan losses, and in an efficient market, Type 2 announcements should consequently have no effect on the share prices, but investors who are prone to overreact may react to such announcements. Table 5 shows that the market reactions to Type 2 announcements are insignificant, which is consistent with efficient markets and no overreaction and in accordance with the results in Bischof et al. (2021). Notice, however, that the results regarding Type 2 announcements are based on only 14 observations.

In summary, there is clear evidence that the financial markets react negatively to an announcement by the FSA of additional loan losses, and that the reaction depends on the severity of the loan loss announcement and the riskiness of the bank. In contrast, there is no evidence that investors overreact to these announcements and hence no evidence that they lead to instability of financial markets.

The second hypothesis is given by:

In Hypothesis 2, we assess contagion effects by examining whether the announcement of additional loan loss reserves in bank A influences the share prices of other banks. The sample includes four banks with halted trading due to bankruptcy/closure, which were excluded from the previous event study but are now included in the contagion analysis.

This paper contributes to the literature on the impact of regulatory disclosures on financial markets. Utilizing Denmark’s unique regulatory environment, where the results of ordinary bank inspections by the Danish FSA are publicly disclosed, we provide new insights into the market and contagion effects of such disclosures. Unlike prior studies that focus on one-time supervisory reports or formal enforcement actions, our research uniquely examines the ongoing disclosures of ordinary inspection results. We specifically focus on inspection reports in which the FSA announces that a specific bank needs to increase its loan loss reserves – one of the key variables in assessing the performance and risk of banks. This focus on individual bank’s asset quality is a common feature with the scarce previous research on stress-tests or formal enforcement actions, but stock markets may react differently to the adjustment of loan loss provisions in our study due to differences in the timing, specificity, predictability, severity and urgency of the information. There are several previous studies based on banks' own announcements of increases in loan loss reserves, but the conflicting results in these studies may stem from the fact that they depend on bank managers' disclosure decisions. In our study, the announcements come from an authoritative external source and are therefore independent of bank managers' own disclosure decisions.

First, we establish whether the announcements by the FSA of increases in a bank’s loan loss reserves have an effect on the share price of the restating bank, i.e., whether the announcements contain value-relevant news to the market. We find clear and strong evidence of negative returns over the period 2009–2020, and we confirm that the market reactions are related to the size of the required additional loan loss provisions. Also, we find no evidence to indicate that investors overreact to these announcements.

Next, we examine contagion effects. We find that the stability of the financial system is not compromised, as large or Systemically Important Financial Institutions (SIFI) banks do not experience contagion effects from FSA announcements regarding other large or small banks. While there is a statistically significant negative contagion effect among small banks, its economic consequences are limited. Additionally, contagion effects are observed from FSA announcements regarding high-risk banks to other high-risk banks, with a minor economic impact. Importantly, our study does not find evidence suggesting irrational contagion effects, supporting the conclusion that the FSA should not delay or refrain from publishing inspection reports disclosing required additional loan loss reserves.

Overall, our findings show that the public disclosure of inspection reports by bank supervisors that include proprietary information about loan losses has the potential to enrich the information environment and consequently increase market efficiency without harming the stability of the financial system. Hence, based solely on immediate stock market effects our results support ongoing, transparent supervisory disclosures about the banks’ asset quality. In other words, our findings support the use of accounting disclosures as a policy tool for bank supervisors.

Our study is subject to certain limitations.⁸ Firstly, the ordinary inspection reports in the study may contain value-relevant information in addition to the required information about insufficient loan loss reserves. However, loan loss reserves are a primary focus in almost all ordinary inspection reports and our paper shows a link between share price reactions and the magnitude of insufficient loan loss reserves, suggesting that this prominent information significantly influences the market response.

Secondly, the study exclusively concentrates on immediate stock market effects, omitting for example considerations for potential reactions from depositors or money market investors withdrawing funds, which may be unpredictable for equity investors and, therefore, unlikely to be entirely reflected in stock price reactions on the announcement date. It is also a limitation of our study that we do not address any consequences of supervisory disclosures that share prices fail to capture. For example, Goldstein and Sapra (2014) argue that public disclosure could potentially enhance the quality of the supervisions, because the work of the supervisors and the way they react would be subject to greater scrutiny and discussions by outsiders, i.e., it could discipline the supervisors. Results in Kleymenova and Tomy (2022) indeed suggest that supervisors take into account the public perceptions of their actions by showing that their intervention is both faster and more frequent in countries with higher news circulation. In contrast, Morrison and White (2013) show analytically that regulatory transparency may not always be socially desirable. It depends on the regulator’s initial reputation. A bank failure may undermine a regulator’s reputation, and except if the regulator's reputation is initially very low, secret bailouts, which do not destroy the regulator’s reputation, may be better. Otherwise, confidence in other banks screened by the same regulator may be lost, which may trigger investors to withdraw their funds and hence end in a situation with financial contagion. Our study does not consider this. We do not consider the effects that mandatory disclosures may have on voluntary disclosures either. Findings in Bischof and Daske (2013), for example, suggest that regulators’ (one-time) disclosure of banks risk exposures induces an increase in voluntary disclosures in the following period. Bank regulators’ openness thus seems to shift the voluntary disclosure equilibrium, which has a positive effect on the whole information environment. On the negative side, Goldstein and Sapra (2014) also mention that the additional public information disclosed by the regulators may crowd out market participants’ use of other information sources, and the increased market sensitivity could magnify any noise in the public information, which may lead to bank runs driven by coordination failures and inefficient investment decisions. If bank supervisors use share prices in their intervention decisions (Novotny-Farkas 2016), then the whole supervisory process may be harmed. Our study does not consider such issues.

Finally, a caveat is about the generalizability of the results. S&P’s list of the largest European banks in 2023 (S&P 2023) shows that the assets of the largest Danish bank—Danske Bank—constitute less than 20 percent of the assets of Europe’s largest bank HSBC. It would certainly be more damaging to the stability of the European banking system if HSBC failed. There is consequently a risk that open inspection reports from the supervisory authorities with information about insufficient loan loss reserves in the most important banks in Europe could lead to larger contagion effects than those found in our paper. However, three banks in our study are represented in the S&P list among the top 40 largest banks in Europe, and the assets of Danske Bank are almost 50 percent of the 10th largest bank in Europe. This suggests that our sample contains some banks that are of some importance in a European context, and they are certainly very important in the Danish context that we investigate. This mix of large and small banks is also found in a number of other European countries. Hence, while we expect that our results extend to at least some other European countries, we acknowledge that this may not be the case in countries with very large and systemically important banks that are crucial to the stability of the European banking system.

Based on a comprehensive review of the literature on bank’s financial accounting, Beatty and Liao (2014) and Bushman (2014) make a call for empirical research that can reveal insights into when, where and how transparency positively or negatively affects banks and the banking system. Similarly, after reviewing US and international empirical literature on the economic consequences of disclosure and financial reporting regulation, Leuz and Wysocki (2016) conclude that “we lack evidence that reporting standards and disclosure regulation produce information spillovers, externalities, and/or network effects.” Despite the caveats, our study adds a piece to the puzzle about the consequences of supervisory disclosures that to the best of our knowledge has not been investigated and found before.