IFRS Practices and the Persistence of Accounting System Classification

Classification is a fundamental process in the better understanding of phenomena in many disciplines. For example, the Mendeleev chart and the Linnaean classification are central to chemistry and biology, respectively. Classifications can also be found in the study of languages (e.g., Bloomfield, 1935), law (e.g., David and Brierley, 1985), economics (e.g., Neuberger and Duffy, 1976) and politics (e.g., Shils, 1966).

In accounting, classifications of countries can be traced back at least as far as Hatfield's work in 1911, and they became a major feature of the accounting literature for some years from the late 1960s (see below). Many classifications show ‘Anglo’ countries together and certain continental European countries together.

This paper asks whether that dichotomous classification (such as in Figure 1) is still discernible in the IFRS practices of large listed companies. IFRS offers considerable scope for companies to choose accounting policies, and therefore it allows national profiles of IFRS practice to emerge. For five major countries (Australia, France, Germany, Spain and the U.K.), these profiles have been reported in the literature (Kvaal and Nobes, 2010).

This paper contributes in three ways. First, it uncovers national profiles of IFRS practices for a further three countries not previously studied in this way (Italy, the Netherlands and Sweden). Then, for the eight countries, a classification of the national profiles is made, revealing that a former two-group classification based on 1980 practices persists. All the eight countries use IFRS, and seven of them are in the EU. That is, after 30 years of harmonization led by the IASC/B¹ and by the EU, international differences are clearly visible and countries form the same groupings as they did decades ago, including an Anglo group that contains Australia (not in the EU) and the U.K. (in the EU).

The third contribution is to show that previous classifications (none of which was based on actual accounting practices of companies, as will be explained) may have been unreliable guides to the accounting differences that matter. This paper classifies reporting in eight countries where the rules are the same, but yet the practices are greatly different.

This paper does not suggest that no harmonization has occurred. However, it contributes further evidence that accounting differences generally are very deep-seated and resistant to harmonization over long periods. There is a major policy implication from this, related to the limited success of the whole IFRS project that was aimed at improving the comparability of financial statements.

LITERATURE REVIEW

Hatfield (1911) considers the accounting systems of France, Germany, the U.K. and the U.S. He puts France and Germany together, leading to a three-group classification. Mueller (1967) has four groups, based on proposed major background influences on accounting. He puts the U.S. and the U.K. in one of the groups (in which accounting is seen as an independent discipline). Seidler (1967), Mueller (1968) and AAA (1977) also examine the influences on accounting that might cause country groupings. Later studies create classifications by using data on accounting rules or on opinions of auditors about accounting practices. Such papers include da Costa et al. (1978), Frank (1979), Nair and Frank (1980), Nobes (1983), Doupnik and Salter (1993) and d'Arcy (2001), as discussed further below.

Gray (1988) and Roberts (1995) note that the above classifications can be split into extrinsic (based on influences on accounting) and intrinsic (based on accounting itself). None of these classifications (or any others known to the author)² uses data on actual practices of companies in annual reports. Clearly, the extrinsic ones do not, but even the intrinsic ones use surveys of rules (e.g., d'Arcy, 2001) or assessments, by auditors, of predominant practices in a country (e.g., Doupnik and Salter, 1993) or of a mixture of rules and practices (e.g., Nair and Frank, 1980). Some of the intrinsic classifications are based on inaccurate data on rules or practices (see the discussions in Nobes, 1981, 2004).

The lack of reference to actual practices can be regarded as a serious weakness. For example, IFRS contains an option to re-measure property at fair value whereas U.S. GAAP does not. However, if IFRS companies do not use the option (as is approximately the case),³ the difference between IFRS and U.S. GAAP is of no practical importance to users of financial statements and it does not undermine the comparability of financial statements. By contrast, the U.S. option to measure inventories on a LIFO basis (not an option found in IFRS) is of major practical importance.⁴

Suppose that Country A allows re-measurement of property but not the use of LIFO, whereas Country B allows LIFO but not re-measurement of property. This would show up as lack of harmony if rules are assessed to measure it, but that would be misleading if no company in either country actually used LIFO or re-measurement of property. By contrast, suppose that both Countries C and D allow options on both the above topics. This would be measured as de jure harmony, although investigation of practices might show that most companies in Country C use FIFO and fair value, whereas most companies in Country D use LIFO and historical cost. In conclusion, classifications and measures of harmony or of harmonization are more likely to be useful if based on practices rather than on rules unless, for some reason, the purpose is specifically to examine the rules.

In this paper, unlike all those above, classification is based on the observation of actual practices in annual reports. For the eight countries examined, the rules are identical,⁵ so there is complete de jure harmony; no differences would be measured if rules were being assessed and no classification would be possible. However, as will be seen, there are large differences in practices, and an interesting classification results.

Jaafar and McLeay (2007) warn of another problem that researchers into classification or harmonization have not adjusted for, or even mentioned. That is, some apparent lack of harmony in accounting practices (within or between countries) may be justifiable because it is caused by differences in underlying economic transactions and therefore the different accounting practices might not hamper comparability. Jaafar and McLeay (p. 158) give the example of inventory accounting, where the choice of FIFO or LIFO might be related to particular real usage of inventory, and that might vary by sector of the economy. On investigation of this and two other accounting topics (depreciation method and goodwill treatment), they find some small effect of sector, but it is greatly outweighed by the effect of country.

None of the above classifications mentions this sector issue. In this paper, I take account of it to some extent, as will be seen, by deleting financial companies where the sector effect is most likely. I also present the sectoral mix of the sample for inspection by readers, and note that previous research on similar data to that used here found no significant sector effect.

One sustained debate in the area concerns whether the two-group classification (Anglo versus continental European) can be substantiated. Cairns (1997), Alexander and Archer (2000) and d'Arcy (2001) dispute the dichotomy, whereas Mueller (1967), Nobes (1983), and Doupnik and Salter (1993) support it. A summary of reasons suggested for why those who dispute the two-group classification fail to find it is as follows: they concentrate on non-representative accounting (i.e., on the consolidated statements of a few large companies in continental Europe),⁶ or they concentrate on the regulatory system rather than on accounting practices,⁷ or they use erroneous data.⁸

Despite the doubts of some of the theorists, much empirical literature on other accounting topics includes a two-group classification as an independent variable, in many cases based on the related common law/code law split (e.g., Guenther and Young, 2000; Hung, 2000; Ali and Hwang, 2000; Ball et al., 2000; Hope, 2003; Barniv et al., 2005).

A note on the Netherlands is needed because it has created problems for classifiers. Da Costa et al. (1978) find that the Netherlands is unusual and cannot be classified. Frank (1979) puts the Netherlands in a U.S. group, but Nair and Frank (1980) using a sub-set of the same data put the Netherlands in a U.K. group. Figure 1 (which is from Nobes, 1983) shows the Netherlands as an outlier of the Anglo group. Parker (1991, p. 229) describes Dutch accounting as ‘sui generis’. D'Arcy (2001), using more recent data, shows the Netherlands in a continental group but outside its core. It is therefore difficult to create a clear hypothesis concerning the position of the Netherlands. The country is usually included with continental Europe in the empirical studies mentioned above.

Nobes (2008) asks whether the old classifications are still relevant in the IFRS world, concluding that they can be for several purposes. First, most accounting⁹ in most countries continues to be based on national rules. Also, other characteristics fit the same classification (e.g. whether countries mandate, allow or ban IFRS for unconsolidated financial statements). One other suggestion was made, which is investigated in this paper: are there national versions of IFRS practice that can be put into groups?

Kvaal and Nobes (2010, hereafter K&N) show that, for five major stock markets (Australia, France, Germany, Spain and the U.K.), the IFRS practices of large listed companies continue pre-IFRS national traditions. Companies can do this because of flexibility within IFRS. The result is the existence of distinct national profiles of IFRS practices. K&N studied 2005/6 financial statements, but Kvaal and Nobes (2012) confirm the persistence of these national profiles through to 2008/9.

HYPOTHESIS AND CHOICE OF COUNTRIES

This paper investigates whether the old two-group classification (as in Figure 1) persists even in the recent IFRS practices of large listed companies. If there is scope for pre-IFRS practice to affect choices of IFRS policies (Nobes, 2006; K&N) and if EU harmonization of accounting had worked well, then one might no longer expect to see EU countries on different sides of a two-group classification of the IFRS practices of their companies. In particular, one might expect the U.K. to be grouped with other EU countries rather than with Australia. However, there are important national forces that might survive both EU and IFRS attempts at harmonization. Nobes (1998) analyses the deep-seated effects on accounting of different financing, tax and legal systems. For example, it is suggested that a country that has few listed companies with widespread ownership will not need financial reporting that focuses on helping users to predict cash flows and will, instead, tend to align financial reporting with tax calculations. Ball (2006, p. 15) notes, concerning IFRS practice, that: ‘the incentives of preparers (managers) and enforcers (auditors, courts, regulators, boards, block shareholders, politicians, analysts, rating agencies, the press) remain primarily local’. Given this, I hypothesize that the two-group classification of Figure 1 will still be found in the 2008/9 accounting policy choices made by companies using IFRS.

The method of choosing the countries for investigation here begins with the fourteen countries of Figure 1. K&N investigated the IFRS practices of companies in the five IFRS-using countries with the largest stock markets. All of these are in Figure 1. Of the other countries in that figure, the lack of IFRS data means that I cannot yet add Canada, Japan and the U.S.A. for this study. I propose to add the next three largest IFRS-using capital markets of Figure 1, that is, Italy, the Netherlands and Sweden.¹⁰ By doing this, I ensure that all seven of the ‘families’ of Figure 1 are represented. The only IFRS-using countries of Figure 1 that are excluded from this study are Belgium, Ireland and New Zealand, which have few¹¹ listed companies of the same size as those examined here for the other countries.

This study, then, includes eight countries. Of these, seven (i.e., all but Australia) are in the European Union (including the largest four of the original six EU members, plus the U.K. which joined in 1973, Spain which joined in 1986, and Sweden which joined in 1995). All seven had implemented the main EU accounting harmonization measures (the Fourth and Seventh Directives on company law) by 1995.¹²

The use of the largest companies and of 2008/9 data should provide the strongest test for the hypothesis, for the following reasons. The expectation is that the largest companies are the most likely to be affected by international influences and therefore the least likely to conform to national traditions. Also, Kvaal and Nobes (2012) show that the influence of nationality on IFRS practices had slightly decreased from 2005/6 (mostly a transition year) to 2008/9. So, if there is a two-group classification for the largest companies in 2008/9, one can have confidence that it would be found for the generality of companies and for 2005/6.

DATA AND METHODOLOGY

K&N studied company choices on all sixteen of the IFRS policy options for which they concluded that the choice was observable. This included nine presentation topics and seven measurement topics. K&N note that some topics are more important than others, but that they all contribute towards answering the question whether pre-IFRS national practices continue under IFRS and whether IFRS practice on any topic is significantly different across countries. K&N do not add the topics together, so weighting is not an issue.

By contrast, for classification, the issue of weighting is important. It is discussed by Nobes (1981) and by d'Arcy (2004). It can be argued that measurement topics are more important than presentation topics. It is proposed here to delete three presentation topics from K&N's list: (a) whether a balance sheet shows an increasing or a decreasing liquidity order, (b) whether an income statement includes a line for ‘operating profit’, and (c) the position of dividends received in a cash flow statement. The remaining thirteen policy topics are shown in Table 1; the first six relating to presentation, and the next seven to measurement. However, I check for robustness by adding back some of these topics, as explained later.

The issue of weighting still remains. All the classification studies use equal weightings because of the difficulty of justifying any other approach. Nevertheless, inspection of the data used in previous papers reveals¹³ that some topics are much less important than others. Here, an attempt has been made to exclude the least important topics, so equal weighting is more defensible.

Nair and Frank (1980) usefully separate measurement practices from disclosure practices. This is more problematic here for the statistical methods because it reduces the number of topics in each of the two categories to fewer than regarded as suitable for separate tests. However, I investigate the results of separating measurement from disclosure under ‘Robustness’ below.

For the five countries studied by K&N, I use the data collected by Kvaal and Nobes (2012) from the 2008/9 financial statements of the largest listed companies. For the additional three countries, I again choose the largest listed companies by taking all the companies in the Financial Times‘Europe 500’ at 31 March 2009. This means 29 Italian companies, 17 Dutch and 26 Swedish. Two Italian companies are then deleted because one used U.S. GAAP and another was a subsidiary. For the remaining total of 70 companies from the three countries, I hand pick data on the IFRS options from the financial reports¹⁴ for the year ended 31 December 2008 or nearest after.

In all, the sample is 287 IFRS financial statements. Table 2 shows the composition of the sample companies by country and sector. Some of the companies are in financial sectors. I follow K&N by including such companies but only for certain policy topics. For example, I exclude financial companies from the data on cash flow statements because these companies have different rules from others. Table 1 shows which topics are excluded for financial companies.

The sectoral distribution recorded in Table 2 shows a broad spread of industries for all the countries. As noted earlier, Jaafar and McLeay (2007) found some small effect of sector on the three topics that they studied, which include one of the topics here (inventory cost determination). K&N asked, for the topics used here, whether the dominance of particular sectors might affect the profile of policy choices in particular countries. Apart from the sector-specific practices of financial companies, they found no effect of sectoral mix on national profiles of IFRS practices. As noted above, I exclude financial companies for those topics for which idiosyncratic practices are apparent.

Frank (1979) and Nair and Frank (1980) use principal component analysis (factor analysis) in order to reveal groupings of countries that have similar accounting according to a database of accounting rules and practices. The results were checked by using multi-dimensional scaling. D'Arcy (2001) uses a different database and applies cluster analysis and produces dendrograms. She also uses multi-dimensional scaling. Gordon (1981) gives a good overview of these various techniques, and there is some further explanation below when discussing results.

All three of the above approaches will be used in this paper. The expectation is that they will all lead to similar results, although presented in different graphical ways. However, if clear differences between the results of the different methods were to emerge, that would be a warning of problems in the data.

RESULTS

Principal Component Analysis

Principal component analysis (sometimes called ‘factor analysis’) processes the data in order to look for components that are selections of practices with different weights that best explain the variance between the objects of study (in this case, countries). Kim and Mueller (1978) and Hutcheson and Sofroniou (1999) set out the procedures.¹⁷ Having identified the principal components, the approach then focuses on those that explain the greatest variance. In particular, it is common to select those that have eigenvalues¹⁸ greater than one. In this case, three such components were identified, explaining 85% of the variance.¹⁹ As an example, component 2 (on which Australia and the U.K. load highly) contains several of the choices of Table 1: a focus on net assets, the use of the SORIE/OCI²⁰ (including for actuarial gains and losses), the use of fair value, and the lack of use of proportional consolidation.

Then, each country is assigned to the component on which it loads the greatest.²¹Table 4 shows the component scores for the countries (using twelve topics).²² As may be seen, there are two groups: continental countries exhibit component 1, Anglo countries component 2, with Sweden as an outlier. The U.K. is nearer to the continental group than Australia is. Germany is the nearest continental country to the Anglo group.

Table 4. PRINCIPAL COMPONENT ANALYSIS SCORES BY COUNTRY

Sampling adequacy is checked by a Kaiser-Meyer-Olkin measure which can take values of 0 to 1. In this case, the score is 0.74, which is fairly good (Hutcheson and Sofroniou, 1999).

Cluster Analysis

Nobes (1983) and d'Arcy (2001) use cluster analysis. Following d'Arcy, I use the method of average linkage between groups. The process first identifies the congruence in policies between each pair of countries. It identifies the most similar pair, in this case Germany and France. It then fuses these two together as a single unit and looks for the next nearest pairing, and so on. The result²³ is shown as Figure 2. The vertical branching lines rise as each new country is added, showing increasing dissimilarity. In this case, Italy, the Netherlands and Spain are gradually added to Germany and France. Meanwhile, Australia and the U.K. form their own pair. Lastly, Sweden joins the fused continental group.

Multidimensional Scaling

Both Frank (1979) and d'Arcy (2001) check their results with multidimensional scaling. This method represents data as a configuration of points in two dimensions. It does not automatically produce clusters but gives a graphical representation of the distances between the countries: ‘When the data have not been forced into clusters, the observer can assess better whether clusters exist’ (Cormack, 1971, p. 340).

Figure 3 shows the result for the ‘modern’ non-metrical solution using two dimensions (Gordon, 1981, ch. 5). Again the two-group classification is clear, with Germany and the Netherlands nearer than other continentals to the Anglo group. Very similar pictures result from the ‘classical’ metric solution. According to the Mardia measure of goodness of fit, 93.0% of the variation is explained²⁴ by the two dimensions.

CONCLUSIONS

Using somewhat informal data,²⁵ based largely on impressions of practices of companies in 1980, Nobes (1983) proposed a classification of the accounting systems of fourteen countries (shown here as Figure 1). The first split of countries was into two groups, with Australia and the U.K. in one group and most continental European countries in the other. That classification was drawn up before any of the EU's directives on accounting had been implemented and before any company in any of the countries considered here had adopted IFRS.

I ask here whether it is possible to discern the same classification after 30 years of harmonization efforts by the EU and the IASC/B. Of the fourteen countries of Figure 1, I remove the three not yet using IFRS, then take those with the eight largest capital markets; meaning that only three small IFRS-using countries are excluded. I then put hand-picked data on 2008/9 IFRS policy choices into a series of statistical classification techniques.

The data used here relate to accounting itself rather than to influences on accounting (e.g., as in Mueller, 1967, 1968) or to regulatory systems (e.g., as in Alexander and Archer, 2000). Further, the data measure accounting practices rather than accounting rules (as in d'Arcy, 2001) or a mixture of rules and opinions on practices (as in da Costa et al., 1978; Frank, 1979; Nair and Frank, 1980). This is the first classification to be based on practices, and the first in the context of IFRS. In this case, it is very clear that the use of practices is more relevant than the use of rules, because all eight countries were using the same rules.

All the statistical techniques lead to the same conclusion: Anglo and continental European groupings can be discerned in the IFRS practices of very large companies. I explain earlier why this is likely to be the case, a fortiori, in earlier periods or for smaller companies. The Netherlands is shown here in the continental group (though not centrally), and it had always been difficult to classify. As in the 1980 classification of Figure 1, Sweden is in the continental group, but noticeably different from the rest.

Some limitations of this study are that I look at only the largest listed companies and only for the latest year available when the data were collected. However, as explained, I believe that this represents the toughest test. Other researchers could extend the study and add more countries. Secondly, there might be some effect of sector rather than country, given that the sectoral mix is not exactly the same for each country. However, research suggests that this is likely to be small. I also admit that equal weighting of topics is arbitrary, although I have eliminated cosmetic topics.

I do not suggest that there has been no harmonization in 30 years. On some major topics, IFRS practice is more standardized than previous international practices were: for example, LIFO has been banned, all subsidiaries are consolidated, and finance leases are capitalized. However, for many topics, national patterns have continued into IFRS practice, and groupings of countries are still in place. As mentioned, I have excluded data on several cosmetic issues that also show these patterns and groupings. I believe that the patterns would also exist for important topics that cannot easily be measured, such as the tendency to recognize impairments or to capitalize development costs. It might be possible for other researchers to gather data on these.

I conclude that the findings add to the evidence that accounting practices flow from deep-seated and long-lasting national influences, so that the practices are resistant even to sustained attempts at international harmonization. If the EU's harmonization efforts had succeeded, one would not expect to see the U.K. still classified with Australia rather than with the other EU countries. This confirms the EU's wisdom in abandoning its harmonization efforts, when it comes to listed companies' consolidated statements, in favour of IFRS. The overall objective²⁶ of the IASB is to increase comparability of financial reporting. This is at its most relevant for the large listed companies, examined in this paper, that are all using IFRS. However, even for them, the paper shows that there are clear country-related differences that would hamper comparability. The IASB's objective includes narrowing differences in practice by changing the regulations. For most of the topics studied here,²⁷ there is no obvious justification, in terms of underlying economic differences, for the variations in practice that result from the choices allowed. The implication is that the continued availability of these choices conflicts with the IASB's objectives.