IFRS reporting, firm-specific information flows, and institutional environments: international evidence

For example, Kim et al. (2012) provide evidence that mandatory IFRS adoption increases audit fees.

The large amount of firm-specific information capitalized into stock prices means that the stock prices are less synchronous with market and/or industry returns. We therefore use the terms higher firm-specific information flow and lower stock price synchronicity interchangeably.

While the European Union (EU) mandated the use of IFRS in the preparation of consolidated financial statements starting in 2005, many other countries are still in the process of converging local GAAP with IFRS. Using a sample of firms from 14 EU countries that were mandated to adopt IFRS starting in 2005, Beuselinck et al. (2009) also find that synchronicity decreases in the year of mandatory adoption, compared with that in the pre-adoption period. However, studying the effect of IFRS adoption in a mandatory setting can create other problems: using a single year (2005) as the benchmark ignores other regulatory changes that can occur simultaneously with mandatory IFRS adoption. In this regard, our sample of both voluntary adopters and non-adopters is less likely to suffer from this problem, since firms decide to voluntarily adopt IFRS in different years. In addition, we find that some of the results provided by Beuselinck et al. (2009) are not easy to interpret. For example, in their Table 4, they find that synchronicity reduces more for the EU Mandatory group than for the EU Late Adopt group, but in all three periods of interest the former group is associated with greater synchronicity than the latter group. The evidence actually implies that mandatory IFRS adoption is not the reason for the decrease in synchronicity, since synchronicity decreases even for the group that has not yet been required to adopt IFRS.

In Veldkamp’s model, information production involves a large amount of fixed cost, and high-demand information has a lower per unit cost of production and thus is available to information users at a lower cost than low-demand information.

The year 1998 is chosen as the starting point because few IFRS adopters were identified before 1998.

Worldscope data field 06010 describes the general industry classification of a specific firm. In our sample, firms in the financial service industries (06010 = 04, 05, or 06) were deleted.

We also measure synchronicity using a single-factor market model in which lagged terms and industry returns in Eq. (1) are excluded and repeat our regression estimations. Though not reported here for brevity (available upon request), we find that the results using this alternative measure are qualitatively identical to those reported in this paper.

Note that the amount of (firm-specific versus common) information being incorporated into stock prices is measured in a relative sense. One can argue that IFRS adoption not only increases the informativeness of firm-specific information but also reduces that of common information. However, our inference on the synchronicity effect of voluntary IFRS adoption is unlikely to change, because in such a case a relatively large amount of firm-specific information than common information is capitalized into stock prices. Admittedly, however, a few studies raise questions about this information-based interpretation of synchronicity and provide evidence suggesting that synchronicity may reflect noises in stock returns that are not related to firm-specific information (Ashbaugh et al. 2006). As in many other studies, our tests are predicated upon the information-based interpretation of synchronicity, given that evidence in support of this interpretation is overwhelming and growing in the contemporary finance (and accounting) literature.

The IASC is the abbreviation of International Accounting Standards Committee and IASB is the abbreviation of International Accounting Standards Board, which succeeded the IASC in 2001.

This approach is conservative, compared with the approach under which firms are classified as non-adopters only if they adopt local standards (i.e., if 07532 = 01), in the sense that our approach is likely to introduce a conservative bias against the rejection of the null hypothesis that IFRS adoption has no effect on synchronicity. As a further analysis (see Sect. 7), we use alternative definitions of IFRS adopters to check whether partial IFRS adoption has an impact on synchronicity and find that the synchronicity impact of partial adoption is smaller than that of full adoption.

The inclusion of observations from these 10 countries with no IFRS adopters into our sample is consistent with previous studies, such as those of Covrig et al. (2007) and Kim et al. (2011). As is further explained in Sect. 7, we re-estimate our main regressions after excluding observations from these 10 countries and find that their exclusion does not alter our statistical inferences on the variables of interest.

For example, Piotroski and Roulstone (2004), who measure synchronicity in the same way as in this study, report a mean and median synchronicity of −1.742 and −1.754, respectively, for their U.S. sample.

Note that the information about the percentage of foreign sales and long-term earnings growth forecasts is firm-specific but publicly available to market participants, including both domestic and foreign investors.

Lennox et al. (2012) point out that, when applying the Heckman two-stage regression, the test for selectivity bias (Lambda) and the endogenized regressor (DIFRS) may be highly correlated, thereby creating a serious multicollinearity problem. However, our test results are unlikely to be driven by this multicollinearity problem: Though not tabulated, we find the variance inflation factors (VIFs) for Lambda and DIFRS to be only 2.86 and 3.40, respectively.

Note that the significantly positive coefficient of Foll is consistent with evidence reported in the U.S. market (Piotroski and Roulstone 2004), in emerging markets (Chan and Hameed 2006), and around the world (Fernandes and Ferreira 2008).

The within-country median (rather than the mean) adjustment is similar in spirit to the procedure used by Kim et al. (2011).

We do not include the inverse Mills ratio (i.e., Lambda) in columns 4 and 5, because the coefficients of Lambda are insignificant across columns 1–3, and the use of the within-country median transformation alleviates concerns over possible endogeneity between DIFRS and Synchronicity.

Exclusion or inclusion of observations in the year of adoption produces statistically similar results.

The Heckman approach addresses concerns over selection based on unobservables, while the PSM approach addresses concerns over selection based on observables.

Specifically, the quartile cut-off points of GDiff are Q1 = 5.5, Q2 = 12, Q3 = 13. GDiff is a country-level variable and the number of firm-year observations is unequally distributed across our sample countries. As a result, sample size could differ across four subsamples that are constructed using the GDiff quartile cut-off points (i.e., Q1, Q2, and Q3).

We also check the sensitivity of our results by deleting firms using IAS (07536 = 02). In this case, we define IFRS adopters as firms adopting IFRS (07536 = 23) only, because one can argue that IAS is older than IFRS and may be different. Though not reported here, for brevity, our results using this narrowly defined DIFRS remain qualitatively unchanged: The coefficient of DIFRS is −0.263 with t = −3.83.

In so doing, we exclude three variables, namely, Nind, Cross, GDiff, because we find that their inclusion produces undesirable coefficient estimates for these variables in our firm fixed effect regression. The STATA and SAS results of our firm fixed effect regression including these three variables show that the coefficients of these variables are biased for technical reasons. These biases may stem from the fact that these three variables do not vary much by year, industry, or firm. Specifically, Nind does not vary much across industry years, Cross is constant across firm years, and GDiff does not vary over years.

The full results are available from the authors upon request.

The firm-specific controls included in Eq. (5) are the same as those included in our main regression, with two exceptions: (1) Accr is included only when its interaction with DIFRS is used as an additional test variable and (2) Cross*INST is included in the regression to make our results comparable with those of Fernandes and Ferreira (2008). For country-level controls, we include only GDiff and GDP in Eq. (5) because the other country-level control variables (Nlist, CtySize, and VarGDP) are highly correlated with INST.

In estimating Eq. (5), we delete a firm from the sample if the country-level scores required to measure INST are missing for the country to which the firm belongs. For this reason, the number of firm-years in Table 5 (N = 15,230) is smaller than the number of firm-years in panel B of Table 4 (N = 15,382).

We conjecture that these insignificant coefficients of GDiff may be driven by the high correlation between GDiff and CIFAR.