Expressions of measurement of accounting quality and its applicability at the international level

The first theoretical manifestations of the multifaceted concept of accounting quality based on the notion of earnings management originated during the 1980s and 1990s. For example, ^{Schipper (1989)} defines earnings management as discretionary management intervention in the financial statement reporting process; years later, this conception was complemented by ^{Healy and Wahlen (1999)}, who suggest that such intervention misleads shareholders about the economic performance of the firm. The concept of earnings management was referred to in turn through an indicator appropriate for the way or form of discretionary intervention exercised by the management on the result accounts, this indicator is described as earnings smoothing, which, according to ^{Trueman and Titman (1988)}, is presented when the direction of the firms takes actions to reduce the fluctuations in the reported earnings.

Subsequently, said theoretical statements of accounting quality, seen from the combined effect of the elements of the reporting system on the main accounting variables, were operationalized and developed as measurement expressions (proxies) during the years 2003 to 2008, and, together with these, expressions related to the level of accounting conservatism and the informative role of accounting (value relevance) were used. ^{Leuz et al. (2003)} documented a first approach when elaborating and applying, at an international level, new expressions for the measurement of accounting quality based on the relation of the variability of earnings with the variability of accruals, and the correlation between the latter and operating cash flows, in addition to other measurements; measurements that were subsequently used and complemented by ^{Lang et al. (2003)} and ^{Lang et al. (2006)} to compare the effects of accounting quality among U.S. and non-U.S. listed firms, and between local and foreign listed firms in the U.S. markets with earnings reconciliation, respectively.

These expressions have subsequently become one of the most popular instruments for comparing whether the quality of financial reporting varies with the use of IFRS standards, in addition to other accounting standards such as US GAAP. The first works reported in this line were conducted by ^{Barth et al. (2006)} and ^{Barth et al. (2008)}, who have documented a lower level of earnings management for those firms that apply US GAAP compared to those that apply IAS (now IFRS), and a lower earnings management, greater opportunity in the recognition of losses, and greater value relevance for those firms that voluntarily apply IFRS at the international level, respectively.

Based on these proposals, the accounting literature based on the quality of the financial statements, and especially that related to the effects of the application of IFRS, has developed vertiginously with three measurement alternatives, two of them based on accounting attributes (earnings management and timely losses recognition) and the remaining one based on market attributes (value relevance).

Applying these alternatives, ^{Paananen (2008)} and ^{Paananen and Lin (2009)} conclude that the quality of the financial statements does not improve significantly in Sweden and Germany, respectively, contrariwise, it shows a decrease during the first years of application of the IFRS; for their part, ^{Christensen et al. (2015)}, ^{Günther et al. (2009)} and ^{Lin et al. (2012)} report similar results for German firms forced to apply such standards, suggesting an increase in the level of earnings management and a decrease in value relevance; however, for German firms that voluntarily apply IFRS, the studies conducted by ^{Christensen et al. (2015)} and ^{Günther et al. (2009)} indicate a decrease in the level of earnings smoothing and an increase in the timely incorporation of losses; only ^{Christensen et al. (2015)} reveal an improvement in the value relevance of accounting magnitudes in the capital markets for these firms.

The accounting quality studies were applied simultaneously in other European countries, such as the United Kingdom (^{Iatridis 2008}, 2010), Greece (^{Dimitropoulos et al., 2013}), and Romania (^{Brad et al., 2014}), where there was a decrease in the level of earnings smoothing during the periods of mandatory application of IFRS, in addition to an increase in the opportunity to recognize large losses and an improvement in the informative role of accounting information for the first two cases mentioned. These results are also extendable to the voluntary application of IFRS in UK firms (^{Iatridis 2008}, ²⁰¹⁰), and for the Greek case, according to the study conducted by ^{Iatridis and Rouvolis (2010)}, and show an improvement in the aspects of earnings management and value relevance for the latter firms from the second period of application of IFRS and onwards in relation to the first period of application of these standards.

With respect to multi-country studies of accounting quality, it can be said that different results are obtained, as in the case of ^{Zeghal et al. (2012)}, who document a decrease (in-crease) in market attributes (accounting attributes) for 15 countries of the European Union. Conversely, the study conducted by ^{Devalle et al. (2010)} did not show an improvement in these attributes since earnings smoothing (the opportunity in the recognition of losses) did not decrease (nor increase) significantly for the 5 largest economies of Europe (Spain, Frane, Germany, Italy, and the United Kingdom). In the same direction, the study conducted by ^{Aubert and Grudnitski (2011)} applied on 13 member countries of the European Union did not show any significant increase in value relevance during the post-adoption period; however, ^{Piot et al. (2010)} managed to confirm, for 22 European countries, a decrease in the level of accounting conservatism.

In this order of ideas, when evaluating the quality of the financial statements from 29 countries that obligatorily adopted IFRS during 2005, ^{Ahmed et al. (2013)} indicate a decrease in said quality, evidenced by an increase (decrease) in earnings management (opportunity in the recognition of great losses) with respect to those non-adopting countries.

^{Capkun et al. (2016)} also find evidence of an increase in earnings management for those countries obligated to adopt IFRS since 2005, as well as for those that voluntarily adopted them before that year and made the change to the 2005 IFRS version, suggesting that the version of IFRS issued by the IASB, during the year indicated, allowed the choice of accounting alternatives suitable for managers to earnings management.

The studies of accounting quality under the methodology employed by ^{Barth et al. (2008)} were subsequently extended to American and Eastern countries. Authors such as ^{Cardona Montoya and Cano Morales (2017)}, when evaluating the effect of enforced of IFRS in Canada, indicate an increase in the level of accounting conservatism and in the informative capacity of accounting magnitudes for its capital market. ^{Zhou et al. (2009)}, for their part, find a decrease in the level of earnings management for Chinese firms that apply IFRS with respect to firms that do not, although they do not show greater opportunity to recognize losses. ^{Liu et al. (2011)}, apart from documenting lower earnings management, report an increase in the value relevance of the accounting information for Chinese firms enforced to implement IFRS in 2007, analyzed against their pre-adoption period.

Similar findings were obtained by ^{Temiz and Güleç (2017)}, ^{Abdullah et al. (2017)}, and ^{Fuad and Wijanarto (2017)}, in their studies for listed non-financial firms in Turkey, Malaysia, and Indonesia, respectively, where they report an increase in value relevance for the first case and a decrease in the level of earnings management for the remaining economies during the IFRS application periods with respect to their corresponding previous periods. Finally, ^{Chua et al. (2012)} conclude on an increase in the quality of the financial statements by finding a decrease in the earnings smoothing, an increase in the timely frequency of recognizing losses, and an increase in the informational capacity of accounting for those Australian firms enforced to apply IFRS since 2005.

Regarding the literature on accounting quality under the concept of value relevance, it was developed through alternative methods to verify the informational capacity of accounting magnitudes in capital markets (^{Hellstrom, 2006}), being the models proposed by ^{Bartov et al. (2005)}, ^{Basu (1997)}, and ^{Ohlson (1995)}, the alternatives present in most of the studies applied and related until now. The model proposed by ^{Ohlson (1995)}, for example, proposes that the variations in the share price could be explained by the amounts of the earnings and the book value of the equity reported to the market. ^{Basu (1997)} directs his analysis towards the concept of earnings conservatism under the conception that negative returns (bad news) are incorporated with greater opportunity in the earnings, with respect to the opportunity with which positive returns are incorporated (good news). For their part, ^{Bartov et al. (2005)} propose their measurement by relating the stock market profitability with the earnings for the period and, when applied to a group of German firms, they indicate a greater informational capacity of the earnings obtained from US GAAP with respect to those obtained with the application of IFRS.

Studies on accounting quality in Latin America

During the last decades, the required use of IFRS by most Latin American countries allowed conducting several studies tending to analyze the impact of this accounting framework on the accounting quality, mainly of firms listed in Brazil, Mexico, and Chile, although the findings documented so far have been somewhat contradictory. ^{Vieira et al. (2011)}, for example, using the methodology employed by ^{Barth et al. (2008)}, show a decrease in the level of earnings management and an improvement in the value relevance for those Brazilian firms that partially adopted IFRS during 2008. Conversely, ^{Klann and Beuren (2015)} indicate an increase in the earnings management in the periods of full adoption of IFRS for these same firms during 2010, while ^{Pelucio-Grecco et al. (2014)}, analyzing the accounting quality during the years 2006-2011, conclude that the transition to IFRS restricts the level of earnings management after completing the implementation with respect to the periods prior to such standards. The results obtained in the previous study have recently been reinforced by ^{da Silva et al. (2017)} for the electricity sector, who, when analyzing the financial statements of firms in this sector between 2003 and 2014, report a reduction in the level of earnings smoothing after the adoption of IFRS, concluding on an improvement in the accounting quality seen from a decrease in the level of earnings management.

For the Mexican case, ^{Conesa et al. (2011)}, when analyzing the adaptation of Mexican GAAP to IFRS for the years 1997-2009, suggest a decrease in the level of earnings management; however, ^{Palacios Manzano and Martínez Conesa (2014)} do not suggest a decrease in this level of management since they find that, for the 1997-2008 period, the magnitude of discretionary accruals (a measure of accounting quality) does not decrease with the use of IFRS in Mexican non-financial firms listed on the NYSE.

For their part, ^{Santana et al. (2014)} indicate that Latin American firms report a lower level of accounting quality during the years 2011 and 2012, when comparing the magnitude of discretionary accruals between firms listed in Brazil and Chile with respect to the United Kingdom and Australia, and with respect to France and Germany; although ^{Cardona Montoya (2018)} and ^{Melgarejo (2017)}, analyzing the impact of the application of IFRS during the years 2006 to 2014 and 1997 to 2015, respectively, suggest an increase in accounting quality for 25 Latin and Caribbean countries and for the 5 major Latin economies respectively, showing that the use of IFRS negatively impacts the magnitude of discretionary accruals, and that the latest work identifies a higher level of association between earnings and stock market prices during the period of application of these standards.

Institutional factors and their effectiveness in Latin America

Recent studies in the accounting literature focus their attention on identifying those institutional determinants of the quality of financial reporting and disclosure (^{Ball, 2001}; ^{Ball et al., 2003}; ^{Ball & Shivakumar, 2005}; ^{Burgstahler et al., 2006}; ^{Hail et al., 2010}; ^{Leuz et al., 2003}), in which, in addition to considering IFRS standards as an institutional factor, they recognize that they are not the only element that conditions the behavior of financial reporting practices among countries (^{Leuz & Wysocki, 2016}), especially the behavior at the level of earnings management (^{Houqe et al., 2012}; ^{Leuz et al., 2003}; ^{Reverte, 2008}).

Several studies warn about the presence of possible interdependencies between the same institutional factors, in addition to being all complementary to affect the quality of the financial report (^{Leuz, 2010}; ^{Leuz & Wysocki, 2016}; ^{Wysocki, 2011}); additionally, the literature calls for a response to the high interest of academics and market regulators to identify those institutions that are necessary, sufficient, or more important for the development of capital markets between countries (^{Wysocki, 2011}). Previous studies, such as those conducted by ^{Kothari (2000)} and ^{Ball (2001)}, already revealed other institutions different from the accounting system that could condition the development of economic and capital markets in a country, for example, legal, economic, and regulatory institutions; however, more recent studies identify and process over 35 countries, a wide variety of institutional factors, in addition to the previous ones, others of a sociological and geopolitical type, and that somehow affect the quality of accounting information (^{Isidro et al., 2016}).

Regarding the presence of these institutions in Latin America and the Caribbean, as revealed by multi-country studies, they have marked differences with respect to the institutions present in the rest of the countries (^{Leuz, 2010}); therefore, in line with previous literature, it would affect the quality of accounting information during the mandatory application of IFRS. Precisely by monitoring the degree of effectiveness of some of the institutions present for the Latin region, the following institutions are identified: levels of judicial independence, independence of the board of directors, and protection of the rights of minority shareholders; levels of financing through capital markets and their development; and the degree of effectiveness of enforcement mechanisms on financial reporting and auditing standards.

The level of judicial independence, as a factor of security and support for the property rights of investors (^{La Porta et al., 2004}) is on average significantly lower for countries with a civil- French law tradition (in which all Latin American countries are found) than for other groups of countries with civil law (German and Scandinavian) and common law (Anglo-Saxon) (^{Porta et al., 1998}), which negatively impacts the financial transparency of firms (^{Bushman et al., 2004}), in addition to presenting less conservative accounting practices with respect to the practices of countries with greater judicial independence (^{Bushman & Piotroski, 2006}); more recent studies, such as the one led by ^{Leuz (2010)}, place the Latin American region in the lowest positions of the rule of law index after the Far East and Mediterranean regions in its order, and only above the African region.

With regard to the independence of the board of directors and the level of protection of the rights of minority shareholders, several studies have suggested that levels of earnings management are likely to increase in countries where their firms have fewer independent directors (^{Ebrahim, 2007}; ^{Klein, 2002}; ^{Peasnell et al., 2005}) and show a greater degree of concentration of ownership (^{Reverte, 2008}), respectively. Although ^{La Porta et al. (2000)} describe that those countries with civil law tradition report a lower level of protection of the rights of investors and, therefore, lower quality in corporate governance measures, recently the Global Competitiveness Measurement Index indicates a slight improvement in the mean with respect to the effectiveness of boards of directors, and in the levels of protection of minority interests in most Latin American countries (^{WEF, 2017}); additionally, the index of protection of minority shareholders calculated in the study conducted by ^{Leuz (2010)} reveals a position of the Latin American region above the Mediterranean and Eastern Europe regions.

On the other hand, the levels of financing through the capital markets (as an expression of measuring market size) and the regulation of these markets are considered to have an inverse relationship with the levels of earnings management in line with the studies by ^{Halabi and Yi (2015)} and ^{Reverte (2008)}, respectively. Although Latin American capital markets have been classified as emerging markets (^{Leuz, 2010}), they report, on average, an improvement in both institutions according to the Global Measurement Index for most of their countries (^{WEF, 2017}), an improvement that is documented by ^{Leuz (2010)} in its securities regulation index, being even above the Mediterranean region, Africa, and Eastern Europe. Additionally, the multi-country study conducted by ^{Houqe et al. (2012)}, which includes the 7 main Latin American capital markets, found the aforementioned inverse relationship between the levels of regulation of these markets and earnings management (measured through the magnitude of discretionary accruals) during the periods of application of IFRS.

Finally, regarding the degree of effectiveness of enforcement and control mechanisms over financial reporting standards and auditing standards, ^{Halabi and Yi (2015)} demonstrate an inverse relationship between the former and the magnitude of discretionary accruals during the application of IFRS standards at the multi-country level. Meanwhile, ^{Francis et al. (2003)} suggest a direct relationship between the latter and the quality of financial reporting. And although some studies have reported negative reactions in capital markets with the adoption of IFRS in countries where the first institution is rated as weak (^{Armstrong et al., 2010}), the Latin American region is characterized by improving its positions for both institutions in most of its countries during the last few years, according to the Global Competitiveness Index (^{WEF, 2017}), an improvement that is evidenced by ^{Houqe et al. (2012)} in its multi-country study (which includes several Latin American countries), noting a lower level of earnings management when both institutions are qualified as rigorous.

Definition of the hypotheses

Based on the findings documented in international literature and especially those related to the Latin American region so far, the following hypotheses are put forward:

H1: It is expected that during the IFRS application periods, the level of earnings management will decrease significantly with respect to the local GAAP application periods in the Latin American and Caribbean region.

H2: During IFRS application periods, the opportunity is expected to increase significantly when recognizing losses (timely losses recognition) with respect to local GAAP application periods in the Latin American and Caribbean region.

H3: During IFRS application periods, accounting information is expected to present a significant increase in informational capacity (value relevance) in the Latin American and Caribbean capital markets with respect to local GAAP application periods.

Earnings management (accounting attributes)

Following ^{Barth et al. (2008)}, four (4) indicators are used to measure earnings management, and these are used to test the first of the hypotheses defined (H1); three (3) of the indicators are related to earnings smoothing and the remaining indicator refers to the earnings management towards a target. ^{Barth et al. (2008)} have documented that with lower earnings smoothing and lower earnings management towards a target, the lower the management of the same and, therefore, the higher the accounting quality will be.

Earnings smoothing

The measurement indicators used are: the variance of the change in the net income (ΔNI) (^{Lang et al., 2006}), the ratio of the change variance in the net income to the change variance in operating cash flow (ΔNI/ΔCFO) (^{Lang et al., 2003}), and the Spearman correlation between the accruals (ACCRUALS or ACC) and the operating cash flow CFO, Corr (ACC, CFO) (^{Leuz et al., 2003}).

The residues obtained from their respective regression are taken from all of them and are defined as follows: ΔNI*, (ΔNI*/ΔCFO*), corr (ACC*, CFO*); the decision to take the residual is due to the fact that each indicator is sensitive to a series of factors not related to the general accounting system, which have been included in the controls defined for their respective regression. The residual from each regression is expected to correspond to those variables not contained in the control variables, which would explain part of the changes in the defined indicators and refer to the elements of the accounting system.

The indicators are interpreted as follows:

(ΔNI*): given a small variance in the residual of the change in the net income, this should be evidence of the existence of earnings smoothing (^{Barth et al., 2008}).

(ΔNI*/ΔCFO*): given a lower ratio in the period of IFRS application with regard to the ratio in the period of local GAAP, the presence of earnings smoothing is made evident. Previous papers have documented that firms with greater volatility in their operating cash flow typically have greater volatility in the net income. Therefore, if firms smooth out the earnings, the variance in the residual from the change in net income is expected to be lower than the variance in the residual from the change in operating cash flow, and this translates into a lower ratio between them (^{Lang et al., 2003}).

Corr (ACC*, CFO*): a higher negative correlation between the CFO and accruals is indicative of greater earnings smoothing (earnings management), suggesting that firms use accruals to smooth variability in cash flows (^{Land & Lang, 2002}; ^{Lang et al., 2003}).

The ΔNI* are the residues of the following equation:

Where:

For firm i in year t,

ΔNI = change in net income before extraordinary items divided by total assets at end of year.

SIZE= natural logarithm of total assets at end of year (firm size).

GROWTH= percentage change in sales.

EISSUE= percentage change in total equity.

LEV= total liabilities divided by the book value of equity, both at the end of the period (leverage).

DISSUE= percentage change in total liabilities.

TURN= sales divided by total assets at end of year (asset turnover).

CFO= cash flow from operating activities divided by total assets, both at end of year. AUD= dichotomous variable, takes the value of one (1) if the firms are audited by a Big4 (PWC, KPMG, EY, and DELOITTE) and of zero (0) otherwise.

NUMEX= number of stock exchanges on which the shares of the firm are listed.

XLIST= dichotomous variable, takes values of one (1) if the firm is listed on a U.S. stock market and that market is not its primary stock exchange. YEAR= reporting year of information.

Country= variable indicator, with Argentina being used as reference (Benchmark). The total number of countries in the sample is 20. For the purposes of the econometric model, 19 va-riables are established to avoid the problem of perfect multicollinearity.

Industry= variable indicator, industry classification based on the NACE code provided by the financial information database ORBIS. Agriculture was used as reference (Benchmark). The total industries in the sample is 19; for the purposes of the econometric model 18 variables are established to avoid the problem of perfect multicollinearity.

ΔCFO* is the residual of the following equation:

Where:

ΔCFO = change in operating cash flow divided by total assets at end of year.

All other variables were defined above.

Therefore, the second indicator is based on the residual ratio obtained from Equations 1 and 2, (ΔNI*/ΔCFO*), comparing this ratio between local GAAP and IFRS periods.

CFO* and ACC* are the residuals of the following equations:

Where:

CFO= is the cash flow from operating activities divided by total assets at end of year. ACC= are the accruals divided by the total assets at the end of year. Total accruals are derived from net income after subtraction of operating cash flow (ACC=NI-CFO).

The other variables were defined above.

The correlation of the residues is obtained from Equations 3 and 4 CFO* and ACC* and compared for both periods.

In both equations the independent variable CFO is excluded.

From Equations 1 to 4 the regressions are estimated by separating the observations for the local GAAP period and the IFRS period.

Earnings management towards a target

This expression is operationalized through the frequency of reporting small positive net income and is evidenced by the SPOS coefficient contained in the regression (5) (^{Burgstahler & Dichev, 1997}), a single coefficient is obtained for both periods and with this one, it is examined whether in one period with respect to the other there is a greater probability of earnings management towards small positive net income.

The results of this model are verified through regression under the LOGIT model and all the observations of both periods are grouped.

Where:

Period (0, 1) = is a dichotomous variable, it takes the value of one (1) for observations belonging to the IFRS period and of zero (0) otherwise.

SPOS= takes the value of one (1) if the net income scaled by the total assets at the end of the year is between 0 and 0.01 and of zero (0) otherwise (^{Lang et al., 2003}). A negative coefficient of the SPOS variable indicates that the firm manages earnings towards small positive net income more frequently in the local GAAP period than in the IFRS period, which translates into lower earnings management for the latter period.

The other variables were defined above.

Timely losses recognition (accounting attributes)

The timely losses recognition operates under the assumption that IFRS standards allow accounting policies aimed at recognizing, in a timely manner, losses in the net income, and this in turn is an indication of an improvement in accounting quality.

This expression is operationalized through frequency when recognizing large losses (^{Ball, 2001}), and with this the second of the hypotheses (H2) is tested.

Where:

LNEG (Large negative Net Income) = is a dichotomous variable, it takes the value of one (1) for observations in which the net income scaled by the total assets at the end of the year is less than -0.20, and of zero (0) otherwise. A positive LNEG coefficient indicates that firms recognize large losses more frequently in the IFRS period than in the local GAAP period, which is an indication of higher accounting quality.

The other variables were defined above.

The results of this model are verified through regression under the LOGIT model and all the observations of both periods are grouped.

Value relevance (market attributes)

It is operationalized through three indicators or measurement models: ^{Ohlson (1995)}, ^{Bartov et al. (2005)}, and ^{Basu (1997)}. These models base their analysis on the informational capacity of the accounting magnitudes in the capital markets, allowing them to test the third defined hypothesis (H3).

^{Ohlson’s model (1995)} refers to the explanatory power of accounting earnings per share (EPS) and book value of equity per share (BVEPS) on changes in share price (P); this is in addition to including control variables that have fixed effects on price such as industry, country, and reporting year. Within the equation, the independent variable POST, and its interaction with the variables EPS and BVEPS, is added. The term POST is a dichotomous variable, it takes the value of one (1) for the IFRS period and of zero (0) otherwise. Differences in value relevance between periods are expected to be reflected in the significantly positive coefficient between the terms of interaction with the POST variable, in β₄ and β₅ of the Equation (7).

Considering the periodicity in which accounting information is publicly available, the share price is measured 6 months after the end of the annual accounting period (^{Lang et al., 2003}; ^{Lang et al., 2006}).

In the model by^{Bartov et al. (2005)} it is expected that the announcement of the accounting earnings will affect the share price and, therefore, its profitability-the dependent variable in the equation is profitability (RETURN), while the independent variables are EPS, POST, and their respective interaction term, in addition to adding the variables industry, country, and information reporting year. The difference between value relevance for the two periods analyzed is expected to be reflected in the significantly positive coefficient of the interaction between POST and EPS, β₃ of Equation (8). Profitability (RETURN) is calculated as the natural logarithm of the share price ratio 6 months after the end of the fiscal year on the share price 6 months before the end of the fiscal year, adjusted for dividends (^{Lang et al., 2006}).

The model proposed by^{Basu (1997)} considers the accounting earnings as a function of returns, incorporating within the model the BAD variable and its interaction with the profitability(return) variable; the BAD variable takes the value of one (1) in the periods in which the firms report negative returns and of zero (0) otherwise. The assumption is that the reported accounting earnings is finally reflected in the share price, the matter to be valued is whether news, especially bad news, are incorporated into the accounting earnings in a timely manner and if this is confirmed through the interaction coefficient between returns and the bad news variable.

The POST variable is also incorporated, as well as its interaction with the other variables indicated so far for this model. Based on the assumption that IFRS contain accounting policies that allow the timely incorporation of losses in the net income, the estimated coefficient ₇ of Equation (9) will turn out to be positive and significant. The magnitude of coefficient ₇ is a measure of the incremental sensitivity with which bad news are reflected in the accounting earnings relative to good news during the period of application of IFRS (accounting conservatism).

Where:

(EPS/P)= are the earnings per share deflated by the share price at the beginning of the period.

All other variables were defined above.

Selection and distribution of the sample

Table 1 presents the selection of the initial sample (Panel A) from 5,545 firms reported as active in the ORBIS financial information database, listed during the analysis period of 2006-2016, with at least one annual report under IFRS and from Latin America or the Caribbean (41,602 observations in total); 3,178 firms are excluded, 1,820 for being listed in securities markets other than Latin American or Caribbean (14,786 observations) and 1,358 firms for not having financial statement reporting under local accounting principles (9,086 observations), making up an initial sample of 2,367 firms and with a remaining total of 17,730 observations.

In Panel B (Panel C) of the same table, the conformation of the sample is reported for the tests related to the accounting attributes (market attributes) starting from the observations selected in the initial sample; in Panel B (Panel C), 13,159 (15,163) observations are eliminated, 5,209 (14,384) for lost data, 7.099 (364) for observations that only present financial information under one of the two accounting regimes, and the remaining 851 (415) observations given that they belong to the transition period from the local regulatory framework to IFRS, forming a final sample for Panel B (Panel C) of 734 (379) firms and a remaining total of 4,571 (2,567) observations

Table 2 shows the distribution of the sample by country (Panel A), industry (Panel B), and year of analysis (Panel C), classified according to the type of test: accounting attributes (Sample A) and market attributes (Sample B). In both types of sample contained in Panel A, a greater participation of firms and observations located in Brazil are present, followed by observations in Chile, Argentina, and Peru.

Table 2. Distribution of the sample

Source: own elaboration

In Panel B, for the sample related to accounting attributes (Sample A) there is a greater participation of observations pertaining to the manufacturing industry and energy supply, followed by firms belonging to the financial activities industry; the observations of firms located in the retail and wholesale trade, transportation and storage, and agriculture and livestock sectors represent a high participation of the total of the observations.

For the sample related to market attributes (Sample B), although the observations pertaining to the manufacturing and gas supply sectors represent more than half of the total sample, the commercial, construction, extractive, and communication activities industries have a representative role in the total sample.

Regarding Panel C, there is a greater participation of observations between 2007 and 2009 under Local GAAP, after these years there is an increasing participation of observations under the IFRS accounting regime, being in recent years when most Latin American countries have incorporated into their legislation the mandatory application of IFRS.

Descriptive Statistics

Table 3 includes the descriptive statistics of the test and control variables used in the analysis of the comparability of accounting quality; as presented in said table, there is a significant difference in the mean (median) with respect to the change in net income (ΔNI) between the periods analyzed, with it being much lower during the period of application of IFRS, at the same time, there is a significant decrease in the mean (median) with respect to the variation in operational cash flows (ΔCFO), in addition to a much more negative and significant amount in the mean and in the median with respect to accruals. It may be alluded that part of the decrease in the variation in the net income is not only due to negative variations in operating cash flows, but also to a greater use of accruals.

Regarding the frequency in the report of large losses (LNEG) there is a significant increase in the mean during the period of application of IFRS, suggesting that possibly during this last period large losses are recognized more frequently than those recognized during the pre-adoption periods. Regarding the frequency in the report of small positive net income (SPOS), although concerning the mean there is a decrease during the period of application of IFRS, it is not statistically significant, suggesting the same level of frequency when reporting small positive net income between both periods.

The statistics also indicate a significant decrease in the mean (median) with respect to the return on assets ratio (TURN) for the period of application of IFRS, this decrease is explained by a significant decrease in the mean (median) growth of firms, analyzed through the variation in sales (GROWTH) and also explained by a significant growth in the mean (median) with respect to the size of the same measured through assets (SIZE).

A significant decrease is reported in the forms of financing of the firm (in mean and median) given the variation in equity (EISSUE) and the variation in liabilities (DISSUE); however, although the leverage (LEV) registers an increase in the mean (median) for the period of application of IFRS, it only turns out to be significant with respect to its median; the above findings do not suggest a relationship between firm growth (SIZE) and the way firms are financed. The increase in firm size during the IFRS period can be explained by the impact of the revaluation contained in the IFRS, and ratified with the same level in mean and median (without statistical significance) of the operating cash flows (CFO) of the firm.

Statistics also report that during the period of implementation of IFRS, firms sustain significant growth in the use of audit services by BIG4 (AUD) and in the acceptance of their listing on a greater number of stock markets (NUMEX). However, the results do not suggest any change with respect to the acceptance of their listing on U.S. markets (XLIST).

With regard to the variables used to measure value relevance, IFRS impact the amount of the earnings per share (EPS) and the book value of equity per share (BVEPS), decreasing significantly (in mean and median) the former and increasing significantly (in median) the amount of equity; the price per share (P) shows a significant fall in the mean (median) for the IFRS period, being extensible to return per share, although the decrease in the latter does not prove to be statistically significant.

Although the normality assumption is not confirmed by the Shapiro-Wilk test, this assumption is relaxed because the sample size is large enough (greater than 30 observations), so normality is assumed according to the Central Limit Theorem.

A t-test is used for the comparison of means between variables, except for those variables in which previously their comparison of variances turned out to be different from zero (significant), in these cases a t-test of comparison of means with different variances is used.

In all variables of the regressions, with the exception of the dichotomous variables, the effects of the outlier values (winsorized) in the 5th and 95th percentiles are controlled.

Sensitivity analysis

Bearing in mind that more than 60% and 70% of the observations in Samples A and B, respectively, belong to firms domiciled and listed in Brazil and Chile, the accounting quality measurement indicators for each of these nations are applied again². For the Brazilian case, the results obtained in aggregate terms do not differ from those reported in Table 4. Never-theless, it should be noted that the informative capacity of the earnings and the book value of equity in the Brazilian capital market (Ohlson and Bartov Models) decreases significantly during the period of application of IFRS.

When considering the Chilean case, although the results do not differ from those obtained in panels B to D of Table 4, they do indicate a significant increase in the level of earnings management given that two of the three indicators related to the level of earnings smoothing turn out to be negative and statistically significant. Finally, addressing the rest of the Latin American and Caribbean countries, the results obtained do not differ from the conclusions reached from Table 4. Likewise, the results contribute additionally to the improvement in the accounting attributes for these countries given a significant increase in the frequency of recognizing large losses (LNEG) during the period of application of IFRS.