Field Conditions and Identification

decisions, such as corporates’ or subsidiaries’ investment cash flow sen-sitivities (Chenet al., 2013; Shroff et al., 2014), and (2) their corporate governance, such as incentive schemes in executive compensation (e.g., Ozkan et al., 2012) or managerial dismissals (Wu and Zhang, 2009, 2019).

Finally, in line with cost-benefit analyses, research has investigated IFRS reporting costs, often proxied by audit fees (De George et al., 2013; Kim et al., 2012). We discuss potential issues with these outcome variables along with the evidence in Subsection 3.4.

Second, for multiple reasons, the treatment itself is not as clean and uniform as the use of a binary indicator variable for IFRS adoption suggests. Most studies compare narrow windows of two years before and after IFRS adoption. This period comprises a number of special effects. In the adoption year, first-time adoption choices are applicable under IFRS 1 (First-time Adoption of International Financial Reporting Standards), reconciliation requirements apply for prior local GAAP numbers, and managers tend to issue transition guidance. The last local GAAP year is special, as firms start to provide outlooks for the likely impact of IFRS adoption, thus becoming subject to the IFRS treatment.

For example, in Australia, firms were even required to reconcile their Australian GAAP to equivalent IFRS figures in the year before full IFRS adoption (Loyeunget al.,2011). In addition, after the European Union’s announcement in 2002 that it would mandate IFRS in 2005, firms started to use their local accounting choices to move closer to IFRS to mitigate severe reporting changes in the transition year. Finally, the IFRS treatment is not uniform across treated countries (jurisdictions often endorse IFRS, instead of mandating the standards as issued by the IASB; see Subsection5.1) and over time (because IFRS are a dynamic set of standards; see Subsection 2.2).

Third, selection bias exists in the timing of the treatment. For exam-ple, many EU firms voluntarily adopted IFRS well before the mandate was even announced. These “early voluntary” adopters (Daske et al., 2008) therefore purposely chose to adopt, and they make the estimation of the IFRS effect subject to self-selection and endogeneity issues. Given that “late voluntary” (or “early mandatory”) adoption was possible between the EU’s announcement in 2002 and the entry-into-force date in 2005, the mandatory treatment of firms around 2005 still contains choice, as these treated firms chose to adopt IFRS late. Furthermore, even the mandatory adoption decision by the legislature cannot be considered to be entirely exogenous. Rather, it was an endogenous choice determined, for example, by perceived economic network benefits (Ramanna and Sletten,2014).

Fourth, natural control groups are missing in practice. Firms must apply IFRS at the level of their jurisdiction, and the mandate often

applies to all listed firms on regulated markets.²² This impedes the possibility of setting up a control group within the same jurisdiction.²³ As a consequence, control groups are most likely to consist of listed firms from non-adoption countries, which are often dominated by a few jurisdictions (the United States and Japan) and require adequate controls for country-specific time trends.

Fifth, there are many other institutional changes that occurred around the same time and make it difficult to isolate the effects of IFRS adoption. IFRS adoption was often not a singular event but part of concerted efforts to strengthen capital market regulation and the financial infrastructure more generally.²⁴Thus, the timing of these other reforms often overlaps with the timing of IFRS adoption, which likely complement each other as their regulatory goals are often very similar.

These confounding events inhibit singling out IFRS effects and represent a core limitation.

3.2.2 Specific Settings

Given the limitations of cross-country settings, researchers have used specific settings that offer conditions that allow mitigating at least

22In some settings, there have been temporary exemptions. EU firms could, for example, delay IFRS adoption until 2007 if they were filing U.S. GAAP reports or if they had only issued debt. There even exist permanent exemptions that may apply to very specific groups of listed firms, such as standalone entities not required to prepare any consolidated financial statements (Pownall and Wieczynska,2018) or firms listed on unregulated (grey) markets, such as the Alternative Investment Market in London (Gerakoset al.,2013). The number of these firms is typically low, and their characteristics are unique, such that any parallel-trends assumption is hard to justify.

23Private firms are exempt from the (full) IFRS mandate in most countries. Private firms have only rarely been used as control group (e.g., Cascino and Gassen,2015), because of the lack of market-based outcomes, and differing reporting incentives (e.g., Bonacchiet al.,2019; Burgstahleret al.,2006).

24For example, the Reports on the Observance of Standards and Codes (ROSC) by the World Bank and the International Monetary Fund (IMF) document how the two institutions promote international standards for 12 different fields of regulation, only one of them being the adoption of IFRS in the field of financial reporting regulation.

some of the described challenges and thus moving closer toward a clean identification of an IFRS effect.²⁵

First, studies have made use of reconciliation requirements in the adoption year, holding the firm and the economics of the reporting period constant. In the initial adoption year, firms are required by IFRS 1 to prepare both the current year’s and the prior year’s financial statements under IFRS (to ease comparison). This requirement allows benchmarking IFRS with the local GAAP numbers disclosed in the prior year, holding the underlying transactions constant (e.g., Hung and Subramanyam, 2007). In addition, IFRS 1 mandates reconciliations of earnings and book values in the footnotes of the first IFRS report, such that users can understand which standards drove material differences (e.g., Barth et al.,2014). In the United Kingdom, firms disclosed standalone IFRS reconciliation documents separately from other disclosures (in timing and content), thus enabling the use of an event-study design (e.g., Christensen et al., 2009; Horton and Serafeim, 2010). However, all reconciliation settings share the limitation of allowing only for cross-sectional analyses in the IFRS adoption year.

Second, studies have made use ofpersistent reconciliation require-ments for firms cross-listed in the United States, holding the firm and the economics of the reporting period constant. Until 2007, the SEC required foreign firms that cross-listed securities on a regulated U.S.

securities exchange to reconcile earnings and book values from their home-country GAAP (often IFRS) to U.S. GAAP. Yearly reconcilia-tions provided researchers with the opportunity to compare summary measures across two accounting regimes for the same firm and year and thus benchmark IFRS and U.S. GAAP (e.g., Chen and Sami,2008, 2013; Harris and Muller, 1999; Henry et al., 2009). However, many studies document that firms that self-select to list shares in the United States are of a specific type, which is neither fully comparable to peers at home because cross-listed firms are subject to U.S. oversight (e.g., Langet al.,2003), nor to U.S. peers, because cross-listed firms remain

25While these settings offer better internal validity of the research design, external validity of the findings is a concern, since IFRS likely interact with institutions that are different elsewhere.

subject to domestic reporting incentives (e.g., Langet al.,2006; Leuz, 2006; Lundholm et al.,2014).

Third, studies have usedsettings where IFRS does not materially change accounting rules, holding the reporting quality (likely) constant.

Brochet et al. (2013) argue that UK accounting standards are often perceived to resemble IFRS. Thus any capital market effects for UK firms after IFRS adoption are likely caused by positive spillovers from other (non-UK) peers adopting IFRS. They back their argument by the observation that Bae et al.(2008) categorize the United Kingdom as having the least number of differences across accounting standards (only one out of 21 differences that they code). However, to the contrary, Barthet al.(2014) report that firms’actual reconciliations of net income from UK GAAP to IFRS reported in the IFRS adoption year tend to be even larger for UK firms than for firms from many other European countries.

Fourth, studies have usedsettings where some firms had voluntarily adopted IFRS before the mandate, holding the reporting rules constant.

For these firms, any observable change in outcomes around the IFRS mandate should be caused by (1) contemporaneous improvements in the set of IFRS standards, that is, new pronouncements becoming effective within the mandatory reporting year;²⁶ (2) spillovers from peer firms that started to report under IFRS; and (3) contemporaneous institutional changes that complemented the IFRS mandate. However, these potential drivers are difficult to disentangle. Still, these voluntary adopters were frequently used as a control group (e.g., Byard et al., 2011; Wang, 2014).

Fifth, studies have usedsettings in which firms have a choice between IFRS and other accounting standards, holding the institutional environ-ment constant (e.g., the former German New Market or Euronext; see Leuz, 2003; Pownall et al.,2014). Such an option allows for the com-parison of different accounting regimes in a single home-market setting (unlike U.S. cross-listings). However, if the choice between standards (mostly between IFRS and U.S. GAAP) is not random, self-selection

26For example, in 2005, four new pronouncements (IFRS 2, IFRS 4, IFRS 5, and IFRIC 2) were added to IFRS, and 19 revised pronouncements became binding for periods beginning on or after January 1, 2005.

is hard to control for, and the literature has struggled to introduce convincing instruments.

Sixth, studies usedsettings in which decision-makers are affected dif-ferently by the IFRS mandate. In this case, the analysis is performed at the decision-maker level. For example, Hortonet al. (2013) distinguish between (1) analysts covering only firms that used a single local GAAP before IFRS adoption, though some of these firms use IFRS and some use local GAAP afterward (for them, comparabilitydecreased); (2) ana-lysts covering only firms that used a single local GAAP before adoption, and all used IFRS afterward (for them, comparability wasunchanged);

and (3) analysts covering firms that used different local GAAPs be-fore adoption, but all used IFRS afterward (for them, comparability increased). Thus, IFRS treatment leads to different consequences de-pending on coverage before the IFRS mandate, and identification rests on the extent to which coverage allocation was exogenous. Comparable identification strategies have been applied to loan syndicates (Brown, 2016) and institutional investors (Covrig et al., 2007; DeFondet al., 2011).