Introduction

According to the ^{IASB (2015)} conceptual framework, the general objective of financial reporting is to provide useful information. To achieve this goal financial reporting must include information about risks and uncertainties faced by companies. Otherwise existing and potential investors, lenders and other stakeholders will not be able to properly assess the risk profile of companies that is necessary for the decision making process (^{Cea García, 1995}; ^{ICAEW, 1999}).

In the last decades there was an increase of research about corporate risk disclosure (^{Abraham & Cox, 2007}; ^{Allini, Manes, Francesca, & Hussainey, 2016}; ^{Madrigal, Guzmán, & Guzmán, 2015}; ^{Linsley & Shrives, 2006}; ^{Oliveira, Rodrigues, & Craig, 2011}); most of them conclude that risk information is not enough (^{Kravet & Muslu, 2013}), triggering doubts about the quality and utility of disclosed risk information (^{Abraham & Shrives, 2014}).

Studies in this field registered different research objectives and methods varying from the identification, by means of manual content analysis, of the nature of risk information included in the narrative sections of firms reports (^{Abraham, Solomon, & Stevenson, 2007}; ^{Abraham & Cox, 2007}; ^{Amran, Bin, & Hassan, 2009}; ^{Atan, Marahun, Kadir, & Jusoff, 2010}; ^{Beretta & Bozzolan, 2004}; ^{Berger & GleiBner, 2006}; ^{Combes-Thuélin, Henneron, & Touron, 2006}; ^{Dobler, Lajili, & Zéghal, 2011}; ^{Lajili & Zéghal, 2005}; ^{Linsley & Shrives, 2006}; ^{Mohobbot, 2005}; ^{Oliveira et al, 2011}; ^{Rajab & Schachler, 2009}; ^{Semper & Beltrán, 2009}) to OLS looking to find firm related variables that may explain the level of risk disclosures (^{Abraham et al., 2007}; ^{Abraham & Cox, 2007}; ^{Amran et al., 2009}; ^{Atan et al., 2010}; ^{Beretta & Bozzolan, 2004}; ^{Dobler et al., 2011}; ^{Madrigal et al., 2015}; ^{Linsley & Shrives, 2006}; ^{Mohobbot, 2005}; ^{Neri, 2010}; ^{Oliveira et al., 2011}; ^{Semper & Beltrán, 2009}; ^{Vandemaele, Vergauwen, & Michiels, 2009}) or assess the utility of disclosed risk information (^{Campbell, Chen, Dhaliwal, Lu, & Steele, 2014}; ^{Kravet & Muslu, 2013}). As in other broad corporate disclosures studies, research in risk disclosure investigated firm specific features, as size, risk, profitability, industry or cross listing.

More recently, under agency, stakeholders or legitimacy theories, due to the relevant role that corporate governance and auditing are expected to play in achieving transparency and avoiding misleading financial reporting, accounting research focused on internal and external corporate governance devices and audit quality as variables that may explain observed financial reporting and disclosure. Following this trend, risk disclosure research has recently investigate corporate governance variables that may explain disclosure (^{Allini et al., 2016}; ^{Al-Maghzom, Hussainey, & Aly, 2016}; ^{Elshandidy & Neri, 2014}; ^{Mokhtar & Mellett, 2013}; ^{Ntim, Lindop, & Thomas, 2013}). However, except for audit size, auditing quality has not, so far, been observed as intensively.

^{Elzahar and Hussainey (2012)} acknowledged that studies on the determinants of risk disclosure have been performed mostly on annual reports and, to the best of our knowledge, the tone of risk disclosure information assessed by the semantic properties of narratives has been studied only in annual reports (^{Dobler et al. 2011}; ^{Linsley & Shrives 2006}). Despite the fact that annual reporting is intended as the main source of information for decision making (^{Solomon, Solomon, Norton, & Joseph, 2000}), and the relevant role it plays in disclosing risks and uncertainties faced by companies, some authors like ^{Hassan and Marston (2010)} propose that the annual report is not sufficient, arguing that other means, such as interim reports, provide more timely information.

Deliberately, we want to carry out in interim reports the kind of research that has been applied in annual reports to observe the "nature of risk disclosures, examining time orientation, whether they are monetarily quantified and if good or bad risk news is disclosed" (^{Linsley & Shrives, 2006, p. 387}).

This paper contributes to filling the gap in the literature about risk disclosure in the interim report and advancing the research of audit quality as a possible determinant of risk disclosure, answering to the call of ^{Elzahar and Hussainey (2012)} for further research on the topic of corporate risk disclosure in interim reports.

The study addresses two main research questions: (i) whether the semantic features of corporate risk disclosures in the interim reports have the same trends that extent literature reveals for annual reports and (ii) whether the level of corporate risk disclosure in the interim reports is explained by audit quality.

Research focuses on risk disclosures made by 35 public Portuguese non-financial companies in its 2014 first half year (1H2014), reports sorted by decreasing market capitalization, using univariate and multivariate statistics on risk disclosure indexes crafted by manual content analysis.

Findings show that quantified risk disclosures are higher in number than unquantified risk disclosures, although the pattern of all others remains the same as in the annual reports. Accordingly, interim report can be considered less boilerplate, which can be explained by the unique nature of this type of report more focused on performance data and with more relaxed disclosure requirements.

Findings also point to no association between proxies of audit quality and risk disclosure indexes, except for Audit Tenure which presents an inverse relationship, indicating that the number of years an auditor stays in a firm is not a condition for better risk disclosure in first half year interim reports. Disclosure policies of sampled firms are not influenced by auditor quality despite the reputational auditor risk that may exist.

The rest of the paper is organized as follows: next section briefly reviews the literature, states the hypothesis for testing and presents the analytical framework. Results, along with their analysis and discussion are presented in a separate section. The last section contains conclusions and suggestions for future research.

Theoretical background and hypothesis

Association studies on corporate risk disclosure

Another area of research in this topic has been carried out in studies that seek to identify, by means of multivariate analysis, associations between the level of risk disclosure, in general also captured by manual content analysis of the annual reports, and some specific variables of companies in order understand which factors may affect the disclosure of risk information by firms. This research has been conducted by different authors, with company size being the factor most widely associated (positively) with risk disclosures indexes (^{Abraham et al., 2007}; ^{Abraham & Cox, 2007}; ^{Amran et al., 2009}; ^{Atan et al., 2010}; ^{Beretta & Bozzolan, 2004}; ^{Dobler et al., 2011}; ^{Linsley & Shrives, 2006}; ^{Mohobbot, 2005}; ^{Neri, 2010}; ^{Oliveira et al., 2011}; ^{Semper & Beltrán, 2009}; ^{Vandemaele et al., 2009}). Profitability was found to be unrelated with corporate risk disclosures (^{Lajili & Zéghal, 2005}; ^{Mohobbot, 2005}; ^{Neri, 2010}; ^{Semper & Beltrán, 2009}), as well as company risk represented by leverage (^{Abraham et al., 2007}; ^{Abraham & Cox, 2007}; ^{Amran et al., 2009}; ^{Atan et al., 2010}; ^{Dobler et al., 2011}; ^{Lajili & Zéghal, 2005}; ^{Linsley & Shrives, 2006}; ^{Mohobbot, 2005}; ^{Rajab & Schachler, 2009}; ^{Semper & Beltrán, 2009}). Few studies have found firm risk associated with risk disclosures (^{Hassan, 2009}; ^{Oliveira et al., 2011}). For industry type research presents mixed conclusions with studies showing no association (^{Abraham & Cox, 2007}; ^{Atan et al., 2010}; ^{Beretta & Bozzolan, 2004}) and others showing association (^{Amran et al., 2009}; ^{Hassan, 2009}; ^{Rajab & Schachler, 2009}). For corporate governance variables, studies usually presents positive association (^{Abraham & Cox, 2007}; ^{Oliveira et al., 2011}; ^{Semper & Beltrán, 2009}; ^{Vandemaele et al., 2009}) as well as for cross listing (^{Abraham & Cox, 2007}; ^{Rajab & Schachler, 2009}).

Our second set of hypotheses relates to the research question of concluding whether corporate risk disclosure could be explained by audit quality.

From an agency (^{Jensen & Meckling, 1976}) and stakeholders (^{Friedman & Miles, 2002}) theories perspective, financial information is crucial as a device to reduce asymmetries. The better the financial information is, the better financial markets and economies perform. Auditing, enhancing the credibility of financial information, is expected to play an important role in the proper functioning of capital markets (^{Francis, 2004}).

Audit quality may be defined by as the joint probability of the auditor finding a material misstatement in financial statements or the accounting system of the audited client and that the auditor will report the discovered material misstatement (^{De Angelo, 1981}). ^{DeFond and Zhang (2014)} defines audit quality as a greater assurance of high financial reporting quality. Also it appears to be influenced by auditor's competence and independence (^{Malek & Saidini, 2013}).

Due to auditor risk reputation that may arise from being associated with clients that adopt unsound reporting practices, auditors have incentives to encourage their clients to disclose comprehensive financial information (^{Clarkson, Ferguson, & Hall, 2003}; ^{Craswell & Taylor, 1992}) including information about risks and uncertainties.

Regulation about the audit of interim reporting vary across countries. In Portugal there is no requirement for listed companies to audit or review interim reporting. Nevertheless, information about risks and uncertainties must be obtained from proper systems of risk management and internal control. One could expect that the quality of external auditors, who must to audit those risk management and internal control systems of companies in an ongoing basis and perform procedures to certify that annual financial statements are free from material misstatements, could be associated with risk disclosure indexes even in interim reporting, considering that reputational incentives exists for auditors to encourage clients to adopt sound financial reporting practices (^{Craswell & Taylor, 1992}) which includes risk disclosure.

Audit quality, being a complex concept, is usually represented by proxies that, according to ^{DeFond and Zhang (2014)}, can be input or output based. Input based proxies refer to observable inputs to the audit process and encompasses auditor specific characteristics, such as auditor size, and audit fees. We also introduce audit tenure as proxy for audit quality.

Auditor size

Large auditors firms (also designed by Big N) usually have higher competences and for that reason they would be able to provide high audit quality to their clients (^{De Angelo, 1981}) as well as better post-audit services. This is due to their capacity of having better training and better facilities. In addition, as auditor reputational risk is higher for large audit firms, it becomes an incentive for recommending their clients to adopt good reporting practices.

Accordingly, it is expected that companies being audited by a Big4 would perform better in risk disclosure, and that motivates our Hypothesis:

H5 - There is a positive association between the number of risk disclosures and being audited by a Big4.

Audit fees

The price auditors charge their clients is ordinarily used as a proxy of audit quality (^{Yang, Yu, Liu, & Wu, 2017}). Audit fees may be seen as the "economic cost of efficient auditors" (^{Salehi, Moradi, & Paiydarmanesh, 2017, p. 37}) and incorporate the expected cost of poor quality earnings (^{Hribar, Kravet, & Wilson, 2014}).

Firms with lower quality financial reporting and disclosure will expect to pay more for the audit service as a result of additional audit hours needed to respond to an assessed high client risk.

^{Shannon, Steven and Nathan (2010)} found that audit fees paid to Big 4 firms increase in moments when risk of misstatements in financial statements due to error or fraud increase. ^{Salehi et al. (2017)} did not find a relationship between audit fees and quality of disclosure while ^{Yang et al. (2017)}, in a work on 10 k fillings using a text mining approach, found a positive association between disclosures of different risks types and audit fees. ^{Hribar et al. (2014)} suggests that while additional audit work may have some impact in improving a firm's accounting quality, is not expectable that could convert a bad financial reporting firm in a good financial reporting firm. Also, it is expectable to find a negative association between unexplained audit fees and accounting quality.

Accordingly, we state our hypothesis as follows:

H6 - There is a negative association between the number of risk disclosures and the value of audit fees

Audit tenure

The selection continuity of auditors may have different impact in audit quality. On one hand, long auditing mandates increases the auditor's experience and knowledge in the client's industry (^{Myers, Myers, & Omer, 2003}) and make it possible to be better familiarized with client business processes and features (^{Carcello, Hermanson, Neal, & Riley, 2002}) improving audit quality. On the other hand, long auditing mandates may undermine independency from the client, and affect audit quality negatively (^{Deis & Giroux, 1992}).

Our hypothesis is as follows:

H7 - There is no association between the number of risk disclosures and the number of years the auditor has in the firm

Analytical framework

Methodology

One single coder (one of the authors) performed manual content analysis over the course of five months, on the narrative sections of interim reports, namely the management report and the notes to financial statements more likely to contain risk information such as provisions, impairments, contingent assets and debts, events occurred after year-end and risk factors when available. The chosen coding unit was the sentence in its context as it is the one that, most likely, allows reliable data to be obtained (^{Milne & Adler, 1999}). Like ^{Linsley and Shrives (2006)}, to ensure reliability, ten reports have been codified twice which made it possible to fine-tune the coding rules defined previously.

A sentence has been codified as risk disclosure if the reader was informed of any opportunity or expectation or any danger, threat, damage or exposure that had or may have an impact on the company, or management of any of these opportunities, expectations, hazards, threats, damages and exposures.

The coding instrument was adapted from ^{Linsley and Shrives (2006)} where risks have been classified in two main categories: financial and non-financial risks. As attributes of disclosure (semantic characteristics), we adopted the quantification (quantified and non-quantified), time orientation (past, future and no time orientation) and impact (good news, bad news or neutral).

Accordingly, 13 risk disclosure indexes have been crafted to be used to test the different hypothesis. With the necessary adaptations, all indexes shown below have the content that we present for the index of total disclosures:

Where, for total risk disclosures:

TRD is the total number of risk disclosures observed.

n is the number total of companies in the sample.

fr _i is the number of risk disclosures observed in company i.

Table 1 shows the risk indexes considered.

Data related with proxies for auditor size and control variables used in the OLS model were gathered directly from interim reports and audit fees was collected from 2014 annual reports (this information is not mandatory in interim reporting). Audit tenure was extracted from the annual report of corporate governance of listed entities in Portugal (^{CMVM, 2014}) and industry type (Industry Classification Benchmark) was gathered from the Euronext's website, resulting in nine different industry types. Once existing literature refers different disclosure patterns for industrial and non-industrial firms (^{Elzahar & Hussainey, 2012}), the nine types were collapsed into two, as follows: Industrial (Oil & Gas, Basic Materials, Industrials, Consumer Goods and Healthcare); Non-industrial (Consumer Services, Telecommunications, Utilities and Technology).

The proxies for audit quality together with control variables widely used in disclosure studies (size, leverage, profitability and industry type) were regressed in an OLS model as follows:

Where:

TRD is the index of total risk disclosures

β₀ - The Intercept.

β₁ ... β₇ - Regression coefficients.

AF is the expenditure with audit fees at year end 2014 for company i.

ATEN is the natural logarithm of the number of years that auditors of company i have in the year end 2014.

AS is a dummy variable that assume the value 1 if the auditor of company i in the year end 2014 is a BIG4 and the value 0 otherwise.

SIZE is the natural logarithm of Total Assets of company i in the 1H2014.

LEV is the ratio Total Debt/Total Assets of company i in the 1H2014.

ROE is Return on Equity calculated in the ratio Net Income/Equity of company i in the 1H2014.

IND is a dummy variable that assumes value 1 if the company i belongs to an industrial activity sector and 0 otherwise.

ε - Error term.

Results: analysis and discussion

Conclusions, limitations and final remarks

The study addresses two main research questions: (i) whether the semantic features of corporate risk disclosures in the interim reports have the same trends that extent literature revels for annual reports and (ii) whether the level of corporate risk disclosure in the interim reports is explained by audit quality.

In relation to our first research question, we can conclude that the interim report does not show the same trends of semantic characteristics of corporate risk disclosure compared to those reported in the literature for annual reports. In fact, quantified risk disclosures are higher in number than unquantified risk disclosures, although the pattern of all others remains the same as in annual reports. Overall, these findings lead to the conclusion that the interim report can be considered less boilerplate. This can be explained by the unique nature of this type of report more focused on performance data and with more relaxed disclosure requirements. There is room for speculation on whether the absence of strong risk disclosure requirements leads managers to not disclose boilerplate.

Regarding the second research question, we did not find the size of the auditor and audit fees as explanatory factors of the amount of risk disclosure on interim reports, while the audit tenure revealed with an inverse relationship. All control variables with the only exception of size were found as not explanatories of risk disclosure indexes. Firms size was found to have a very strong effect on the level of risk disclosures confirming the trend widely observed in prior research (^{Dobler et al., 2011}; ^{Madrigal et al., 2015}; ^{Linsley & Shrives, 2006}; ^{Oliveira et al., 2011}) while neither profitability (ROE) nor Industry type explain risk disclosure. Also, firms risk (leverage) does not explain risk disclosure, thus, neither revealing nor concealing arguments referred to by ^{Dobler (2008)} are favored.

As two proxies of audit quality (Big4 and audit fees) were found to have no explanatory capacity and audit tenure revealed an inverse relation, we have to consider that, in this setting, firms risk disclosure policies are not influenced by auditors quality whether or not reputational auditor risk exists.

Manual content analysis is a very time-consuming process, resulting in smaller samples. This limitation can be overcome in the future using electronic content analysis tools. Also, despite rigor along all the process to achieve reasonable levels of reliability, subjectivity may always play a part. As in all similar empirical studies, these types of constraints have to be considered when interpreting results, setting conclusions and making generalizations. Furthermore comparing results with other empirical studies can be influenced by differences in samples, time periods, data gathering process and statistical methods applied.

This research has several empirical contributions which could be of interest for both academics and regulators. It contributes to fill the gap in research about risk disclosure in the interim report, identifying the nature of corporate risk disclosures and assessing the quality of risk information. It also updates research about determinants of risk disclosure in interim reports, namely observing the role of audit quality.

Considering the scant academic work about risk disclosure in interim reports, we encourage future studies to observe other settings comparing results of different countries and enlarging the scope of the determinants of risk disclosure.