Empirical evidence on the evolution of international earnings.
We document convergence in earnings multiples for a sample of firms from Australia, Canada, France, Germany, Japan, the United Kingdom, and the United States over the years 1987-1999. Convergence persists after including controls for earnings, sales and GDP growth rates, interest rates, and returns, and is robust to inclusion of cash flow multiples. We obtain similar results focusing on accruals multiples, suggesting that the pricing of accruals drives the empirical results. We document that cash flow/accrual correlations have become more similar for our sample, and have become less negative for all countries other than the U.S. and Canada. Finally, we provide evidence that book value multiples have converged, although the evidence is weaker than for earnings. Taken together, our results suggest a reduction in cross-country accounting differences over time.
There are several reasons to expect cross-country convergence in accounting. Most directly, there have been significant efforts to reduce accounting differences and develop consistent regulation worldwide through organizations like the International Accounting Standards Committee (IASC), International Auditing Practices Committee, International Organization of Securities Commissions, European Union, and Association of Southeast Asian Nations. Countries like the U.S., Canada, and the U.K. have worked together and with the IASC to develop joint standards, and auditing firms have been active in forming international alliances to encourage consistent practice across countries.
Further, cross-border competition for capital has created incentives to improve accounting quality and comparability. Anecdotal evidence suggests that some large companies have voluntarily switched accounting principles to be more consistent with international norms, and companies have retained international auditing firms to provide greater confidence in their reports. (1) As a result, accounting practice appears to have become more consistent internationally, leading some to conclude that a continuation of the process currently in place with incentives for firms to voluntarily move toward international norms can result in high-quality, comparable accounting practice going forward. (2) For example, the NYSE and others have argued that relaxed cross-listing standards and increased reliance on IAS may be appropriate if cross-border accounting differences are no longer substantial (especially for issuers that would consider listing in the U.S.) and are getting smaller (Cochrane 1994; Cochrane et al. 1996).
Others view progress as more apparent than real, believe that substantial differences remain, and express concern about the perception that differences in accounting have been substantially reduced. The FASB, for example, tallied 255 differences between IASC standards and U.S. GAAP, and concluded that many important differences still existed internationally (FASB 1996). Similarly, Choi et al. (1999), and others, suggest that important accounting differences remain across countries, although the differences identified are generally in alternatives permitted rather than explicit differences in practice.
In the end, the extent of convergence of financial reporting internationally is an empirical question. Our goal is to provide descriptive evidence on whether accounting differences, as reflected in differences in earnings multiples, appear to have decreased over time and, if so, what factors explain the change. Our primary focus is on earnings multiples because earnings is a key summary performance measure in all of our sample countries and because earnings (as a summary measure) should be strongly affected by changes in accounting practice.
While we view our analysis as descriptive, we are guided by two constructs of interest to regulators: "comparability" and "quality." Both constructs are subject to differences in interpretation and are difficult to operationalize, so we do not attempt to measure them directly. Like much of the literature, we take an indirect approach, based on the valuation characteristics of the resulting net income, to infer changes in underlying accounting practice. What we have in mind for comparability is the idea expressed by Jenkins (1999), that "what we all want and need is a set of high-quality standards that assures that similar transactions are accounted for similarly and that unlike transactions are not." We base our assessment of comparability on the multiples applied to accounting data by investors, and argue that increased comparability will be reflected in more similar multiples, all else equal. Our accounting quality construct is a measure of smoothing based on the correlation of cash flows and accruals; we argue that less smoothing implies increased earnings quality. (3)
Our indirect approach takes into account the effects of actual accounting practices, from the perspective of users of accounting information, by basing analyses on investor valuations of reported earnings. We do not attempt to examine accounting practices directly. The extent to which companies exploit allowable alternatives and the magnitude of the resulting effect on income, while potentially large, is generally not observable, and studies like Ball, Robin, and Wu (2000) and Pownall and Schipper (1999) note that accounting changes often result from incentives rather than regulation. The strength of auditing and enforcement also varies across countries, potentially resulting in differences in the application of accounting standards. Further, it is difficult to judge accounting practice based on accounting policy footnotes since disclosure is typically vague, especially outside the U.S., and covers, at most, major accounting choices (Frost and Ramin 1997).
However, our indirect approach creates the potential for omitted correlated variable bias. To counter that possibility, we include tests based on recent approaches to measure earnings management; we also identify where we expect effects to be most pronounced and in what direction, and attempt to incorporate likely correlated variables into our empirical analysis. Further, systematic differences across countries should be mitigated by the fact that we conduct our comparisons over time.
We interpret our findings as most consistent with the notion that the convergence in earnings multiples reflects, at least in part, convergence in accounting practices over time, but that systematic accounting differences remain. Further, our evidence on the time-series properties of earnings suggests that smoothing has been mitigated, and has become more similar across our sample countries. While we cannot dismiss the possibility of omitted correlated variables, these results are robust to a variety of controls.
In the next section, we describe the data. Then, we present the analysis of changes in earnings multiples, followed by our earnings smoothing tests and supplementary tests based on the multiple applied to book value. We conclude with a summary and discussion of caveats.
Our data are from the 1997 and 2000 Global Vantage disks, which is the most comprehensive dataset with the longest time-series that we could find. (4) Our analysis excludes firms in developing economies because they generally lack a sufficient time-series to permit inference, but includes firms in countries that have historically been at the extremes in terms of reporting philosophy. Further, our sample countries all have developed, diversified economies, so firms should face fairly similar economic environments. We exclude financial firms, because they face unique reporting issues and there are not enough to permit a separate analysis.
Table 1 provides descriptive statistics for our sample firms. In the average year, our sample includes 130 Australian firms, 282 Canadian firms, 168 German firms, 172 French firms, 601 U.K. firms, 902 Japanese firms, and 1,940 U.S. firms. Not surprisingly, the larger economies with a longer history of equity ownership tend to have more firms with available data. Because the number of firms varies across subperiods, we replicated tests limiting the sample to firms with data in both periods. While sample sizes are smaller, especially for some of the smaller countries, results are consistent. Japanese, German, and French firms tend to be larger, while U.K., U.S., Canadian, and Australian firms tend to be smaller, but in each case there is a substantial overlap in the distribution of size across countries. Results that follow are robust to controls for size. (5)
III. CHANGES IN MULTIPLES OVER TIME
Table 1 presents earnings/price ratios by country. Following the conventions in Global Vantage, earnings/price is defined as earnings before extraordinary items divided by market value at year-end. We begin with a potential sample of 62,378 firm years and use the same deletion rules as Joos and Lang (1994). Because earnings/price ratios are difficult to interpret for loss firms, we delete observations with negative earnings (18 percent of potential observations in the first period and 20 percent in the second period) and, to mitigate the effects of extreme observations, we delete the top 1 percent of earnings/price ratios, leaving 50,333 observations. (6)
The overall pattern in earnings/price ratios is generally consistent with that reported in other studies, beginning with Speidell and Bavishi (1992). Earnings/price ratios are lowest in Japan and Germany, consistent with their legal origins as code-law countries with close links between financial reporting and tax reporting, and a historical creditor focus, resulting in more conservative income measurement. (7) Consistent with the differences in multiples reflecting the effects of traditional accounting conservatism, median return on equity for Japan and Germany is the lowest among our sample countries during the 1987-1992 subperiod. Similarly, GDP growth rates (not tabulated) during the sample period are no higher for Japan and Germany than for other sample countries, suggesting that high growth does not explain the higher multiples. Similarly, earnings/price ratios and return on equity are generally higher in the common law countries (Australia, Canada, U.K., and U.S.), consistent with those countries following more equity-focused, less tax-based accounting and, consequently, exhibiting less conservative accounting practices. The ordering across countries is consistent with the Barth and Clinch (1996) finding that Form 20-F net income reconciling items are on average negative for Australia, Canada, and U.K., and largest for Australia.
To provide initial descriptive evidence on changes in earnings multiples over time, we estimate a regression after splitting the sample period into two subperiods, comprising the first six years and the last six years. (8) We regress mean-adjusted earnings/price ratios on 14 indicator variables for each country/subperiod grouping (i.e., seven countries multiplied by two subperiods). We adjust each earnings/price ratio by subtracting the average (over firms, years, and countries) earnings/price ratio for that subperiod (.067 for the first subperiod and .064 for the second subperiod) so that the coefficient estimates in the regression represent deviations for a given country from the mean. (9)
Table 2 contains the regression results. In general, a comparison between periods suggests convergence in the sense that dispersion is reduced. In addition, convergence is apparent for all seven sample countries, with each moving toward the mean in the second period. For example, the mean earnings/price ratio for Australia, which was .0284 greater than the mean in the first period, is .0078 greater than the mean in the second period. Earnings/price multiples in Canada, France, U.K., and U.S. also move from above the mean toward the mean. Finally, multiples in Germany and Japan, which were below the mean in the first period, move upward toward the mean in the second period. An F-test of the difference in coefficient estimates between the first and second periods indicates the shift is significant at the .001 level for each country but Canada. However, multiples in most countries remain significantly different from the mean in the second period, indicating that whatever forces caused divergence in the first period have been reduced but not eliminated. (10)
In general, the results suggest a narrowing of differences in earnings multiples over the sample period for all countries, with results strongest in the traditional code-law countries. Subject to concerns about other potential economic determinants, this result is consistent with the assertion by many commentators that, worldwide, accounting differences have been narrowed and that much of the recent progress in harmonizing accounting has occurred in code-law countries, with standard setters adopting more international norms and companies going beyond national requirements to attract capital. (11) Further, the results are consistent with the concern expressed by some (e.g., Biener 1994) that accounting worldwide has moved toward a common law, equity-focused perspective to the detriment of other stakeholders. Similarly, the evidence on earnings multiples for Australia is consistent with its historic reputation for aggressive accounting and its more recent movement toward IAS (Economist 1999).
An alternative test for the narrowing of differences is to compare the cross-country coefficient of variation of average earnings/price ratios in each year (i.e., seven observations to each variance). (12) Comparing the first six years in the sample period to the last six provides striking evidence of a reduction in dispersion (results not tabulated). Each yearly coefficient of variation in the first period is larger than the corresponding coefficient in the second period. In fact, the smallest of the six first-period annual coefficients of variation is larger than the largest of the six second-period coefficients of variation, suggesting that the relation is consistent and is not driven by a subset of years. Applying a rank sum test comparing the first and second periods indicates that the reduction in average variance is significant at the .01 level. The rank correlation between coefficients of variation and time is -0.87, significant at the .001 level with only 13 observations.
A second alternative approach is to compare the proportion of the variation in multiples that can be explained by country differences across the two periods. When we estimate a regression of earnings/price ratios on the country indicator variables (omitting one country to be captured in the intercept), the [R.sup.2] measures the extent to which country-specific differences explain multiples. In the first period, the [R.sup.2] is .17 vs. .07 in the second period, indicating that country-specific factors are more important in the first period (results not tabulated). The difference in [R.sup.2]s is significant at the .01 level based on a Cramer Z-statistic.
Because earnings multiples are a function of economic factors like expected growth and risk, the results based on these tests may simply reflect convergence in other economic factors. We take several approaches to address this possibility. First, we replicate the preceding analysis for cash flow multiples. To the extent that economic factors and not accounting practices are important, they should also affect the multiples applied to cash flows. Because statements of cash flows are not available for most of our firms during the sample period, we estimate cash from operations using net income adjusted for changes in balance sheet accounts. (13) Results for cash-flows-to-price ratios, reported in Table 3, show no consistent evidence of either convergence or divergence between the two periods. Cash flow multiples for Canada, Germany, and U.S. move away from the average, and Australia and France show no change. The fact that cash flow multiples change between periods is not surprising, since factors like expected growth and risk may change as well. Given the results in Table 3, we include controls for cash flow multiples in the analysis that follows.
Our second approach to controlling for economic effects on earnings multiples is to explicitly control for several factors identified by previous research that might affect earnings multiples. French and Poterba (1991) and Guenther and Young (2000) suggest that differences in earnings multiples might reflect differences in GDP growth or real interest rates. To examine that possibility, we collected data on GDP growth rates from the OECD Historical Statistics and on real interest rates from the IMF World Economic Outlook. Similarly, we gathered earnings and sales growth rates for each firm over the sample period to control for the possibility that convergence in multiples might reflect changes in firm-specific growth. (14) Finally, we collected data on prior and contemporaneous returns to investigate if the change in multiples reflects the effects of past or future returns due to some combination of misvaluation and deviations between forecasted and actual growth.
We first examined the control variables for evidence of patterns over time. To the extent that the pattern in earnings multiples is being driven by convergence in economic factors captured by the control variables, there should be evidence of convergence in GDP, earnings and sales growth rates, interest rates, and returns. Plots of each control variable (not tabulated) in each subperiod provide no evidence of convergence in any of the measures. The dispersion of each of the firm-specific variables increases between periods, suggesting that earnings multiple results are not driven by firms becoming generally more similar over the sample period.
More to the point, including the controls in the regressions does not affect conclusions. Table 4 presents results for mean-adjusted earnings/price ratio regressions with controls included. The number of observations decreases because of missing data on returns or the change in earnings. However, the pattern of results is similar to that in Table 2, with all countries except Canada moving significantly toward the mean. Statistical tests yield similar conclusions as before; based on a rank sum test, the cross-country variance of earnings/ price ratios decreases significantly (at the .01 level) between periods after controlling for these potential economic determinants.
We again compare the explanatory power of country indicators using a two-stage approach. First, we regress earnings/price ratios on the economic controls to capture the effects of these economic factors by period. Second, we regress the residuals from the first stage on the country indicator variables to compare the extent to which country-specific factors explain the residual variation in multiples in the two periods. As before, the difference in [R.sup.2]s (not tabulated) is significant at the .01 level, indicating that, even controlling for several economic variables, the importance of country-specific variables in explaining the dispersion in earnings multiples has decreased over time. While our control variables are likely measured with error, they should be correlated with the "true" economic determinants, yet our results are at least as strong with the controls as without.
Another possibility is that convergence in earnings multiples reflects risk shifts or changes in market segmentation. While we know of no empirical evidence of risk shifts during the sample period, this possibility is difficult to dismiss because the nature of priced risk is not well understood, especially in international contexts. We do not have sufficient returns data to estimate a capital asset pricing model for our early period and it is not clear, even if we did, which pricing model is appropriate for our purposes (a one-factor model using each country's index assumes away any effect since the average beta is constrained to be 1). Following Dhaliwal et al. (1999), we use leverage as our proxy for risk. In general, there is no evidence that leverage became more similar during the sample period and including leverage as a control does not affect our conclusions (results not tabulated). Further, for a subset of firms with monthly returns data on Global Vantage, we estimate betas relative to local and world average returns. While the resulting betas are very noisy due to the small number of observations, there is no evidence of convergence in those estimates and our results are robust to their inclusion.
A related possibility is that earnings multiples are converging simply because the underlying economics of our sample firms are becoming more similar. If that is true, then within-country dispersion should also decrease over time. To examine that possibility, we compare the annual within-country coefficient of variation over time for the sample firms. Because data are available for a much longer time for the U.S. firms and trends may be more apparent over longer periods, we also compare the coefficients of variation over the period from the 1950s to 1990s for U.S. firms. Results from this analysis (not tabulated) provide no evidence of a general reduction in variation over time. Of the seven sample countries, four experienced increases in dispersion over the sample period. The U.S. analysis suggests an increase in dispersion over time, primarily in the late 1970s, although the pattern is basically flat during our sample period, which covers 1987-1999.
The preceding analysis includes variables that previous research has shown are likely to influence our results, but it is possible that potentially important unidentified variables are excluded. To the extent that such unidentified factors affect the earnings multiples, they should also be reflected in the multiples applied to cash flows. As noted above, there is some evidence that cash flow multiples change between periods, so controlling for the cash flow multiple may mitigate the effects of omitted economic factors on earnings multiples. (15) We include the ratio of cash flow to price, interacted with indicator variables for each country/subperiod, to control for changes in cash flow multiples between periods. (16) Changes in expected profits or risk should be at least partially captured in the cash flow multiple, and controlled by the inclusion of cash flows in the regression.
Results including both the economic controls and cash flow multiples are reported in Table 5. The results are very consistent with those reported previously, indicating that controlling for cash flow multiples does not change the conclusion that earning multiples have converged over time. With the exception of France and Canada (which were not significantly different from the mean in the first period), the change in coefficient estimates is significant and the rank sum tests (not tabulated) indicate convergence. While there is substantial evidence of convergence, however, the coefficient estimates in the second period generally remain significantly different from the mean, indicating that convergence is incomplete. Again, we apply a two-stage approach in which we first condition on the economic and cash flow controls and then assess the ability of the country indicator variables to explain remaining variation (results not tabulated). As before, there is strong evidence (significant at the .01 level) that the ability of country-specific factors to explain remaining variation is lower in the second period than in the first.
The results in Tables 3 and 5 indicate that the pattern in earnings multiples is robust to controls for changes in cash flow multiples, suggesting that the pricing of accruals drives the results. We test this implication by replicating the analysis of earnings multiples using accruals deflated by price as the dependent variable, including controls for cash flows and other economic factors. Table 6 reports coefficients on accruals/price regressions analogous to those in Table 5, but with earnings replaced by accruals.
The pattern of convergence in accruals is similar to that for earnings, suggesting that the pricing of accruals is driving the changing multiples on earnings. For all countries except Canada and France, the multiple moves closer to the mean in the second period and, in every case in which the multiple was significantly different from the mean in the first period, the change is statistically significant at the .01 level or better. Further, for the cases in which the first period coefficient is not reliably different from the mean, there is evidence of a significant change between periods only for the U.S., suggesting that changes occur only in cases in which they would be predicted. Finally, in most cases the multiple retains its sign relative to the mean in the second period, indicating that accounting influences on earnings multiples in the first period were mitigated, but not eliminated. Applying the rank sum test described earlier, the reduction in dispersion is statistically significant at the .01 level. Comparing [R.sup.2]s, the ability of cross-country differences to explain variation in the residual after controlling for the other factors is significantly (at the .01 level) lower in the second period than the first.
To provide a better sense for the pattern in reduction in dispersion over time, Figure 1 presents the coefficient of variation of accruals multiples after controlling for cash flows and economic variables (analogous to Table 6) over the sample period. Data are grouped into two-year subperiods to increase the number of observations for each country in each subperiod. The pattern in Figure 1 suggests a fairly consistent reduction over the first two-thirds of the total period. While there is less evidence of a reduction in the last third of the sample period, evidence of convergence is not limited to a subset of periods or to one or two countries. (17)
[FIGURE 1 OMITTED]
While we are generally agnostic as to the cause of the observed convergence in earnings and accruals multiples, we provide additional analyses of three possible causes: changes in consolidation policy over time, an increase in the proportion of our sample firms listed on U.S. markets, or an increase in the number of our sample firms reporting under nondomestic standards. Based on the Global Vantage code for consolidation policy, the proportion of firms filing unconsolidated financial statements decreases from 9.0 percent of sample firms in the first period to 4.5 in the second period. However, results are consistent excluding unconsolidated firms.
The second possible explanation is an increase in the number of sample firms trading on U.S. markets. Although Global Vantage reports data under local GAAP, research like Lang et al. (2002) suggests that firms alter their local-GAAP accounting choices around the time of a cross-listing. To assess the extent to which our sample firms are cross-listed on U.S. markets, we compare our sample with a list of cross-listed firms from Compustat. The proportion of sample firms cross-listed on U.S. markets remains at about 8.0 percent between the two subperiods, and excluding those cross-listed firms does not affect our conclusions.
Finally, it is possible that more sample firms are choosing to report under standards other than local GAAP. Based on Global Vantage reporting codes, the proportion of sample firms reporting under IAS or U.S. GAAP has increased over time, but it averages only 0.8 percent and 2.1 percent in the first and second sample periods, respectively, and the use of IAS or U.S. GAAP is concentrated in Germany and France. Conclusions are not sensitive to excluding those observations. Taken together, these results suggest that changes in consolidation policy, cross-listing, and reporting under nondomestic standards do not explain our empirical results.
IV. CHANGES IN EARNINGS SMOOTHING
Because our analysis so far relies on share price to detect accounting differences, it is possible that our results may be driven by factors like market inefficiency, market segmentation, or inadequate controls for risk or expected growth. In this section, we assess convergence by looking to the time-series properties of earnings, without reference to share price.
A frequently cited worldwide concern is the ability of firms to smooth earnings by setting up reserves in good times to be drawn down later. While reserves are an issue everywhere, the rules regulating them are probably most stringent in the U.S., where they have been an ongoing priority for the SEC. Outside the U.S., especially in code-law countries, there has been less emphasis on regulating reserves, perhaps because understating net assets is viewed as less of a concern than overstating them. Foreign registrants frequently reverse reserves when reconciling their financial statements to U.S. GAAP and firms in Germany and Japan typically report lower earnings variability, consistent with smoothing (Leuz et al. 2001; Ball, Kothari, and Robin 2000).
One approach to assessing the effects of income-smoothing behavior is based on the correlation between accruals and cash flows. Authors like Myers and Skinner (2001), Beatty et al. (2002), and Leuz et al. (2001) have argued that, all else equal, earnings smoothing should induce a more negative correlation between cash flows and accruals. For example, Leuz et al. (2001) find that measures based on this correlation approach are highly correlated with other measures of earnings management, and differ predictably across countries based on the importance of equity markets, ownership concentration, shareholder protection, disclosure, and enforcement. (18)
While there may be intrinsic country-specific economic factors that cause the relation between cash flows and accruals to vary, we know of no reason to expect the underlying economics to drive the cash flow/accruals correlation in any systematic direction, especially within countries over time. Further, research like Leuz et al. (2001) finds cross-country results using this measure that are consistent with predictions.
Table 7 presents results for the correlations over time based on accruals and cash flows as calculated in the previous section, deflated by total assets to mitigate the effect of size. (19) We use rank correlations to reduce potential effects of extreme observations, but results are consistent for Pearson correlations. In general, the cross-country ordering of correlations is consistent with results in prior research. Germany, Japan, and France have the most negative correlations, suggesting that earnings smoothing is more pervasive in those environments. Further, the magnitudes are striking, with correlations of about -90 percent in each case. Correlations are less negative, but still quite large, in the U.K. and Australia. The U.S. exhibits the smallest correlation at -56 percent with Canada close behind at -66 percent, suggesting that earnings smoothing in the U.S. is less pronounced than in the other leading market economies. (20)
Most countries experience a significant reduction in the negative correlation between cash flows and accruals between the first and second sample periods. Following Beatty et al. (2002), we test for the significance of differences by estimating by-firm correlations in each subperiod and conducting a rank sum test across the two periods. There is a significant decrease (at the .01 level) in the correlation magnitude for each country except the U.S. and Canada. (21) The fact that evidence of a change is weakest in the U.S. and Canada, where the pricing evidence was also weakest, suggests that the two sets of tests may reflect the same underlying phenomenon. The U.S. is the only country that experienced a more negative correlation in the second period, although the difference is small (-.565 vs. -.557, Z-statistic = 1.86). Our results also suggest that predictable differences remain across countries, with Germany, Japan, and France showing the greatest evidence of smoothing, even in the second subperiod. (22)
V. OTHER ANALYSIS
Our primary focus is on earnings since this measure has attracted the most attention from regulators and analysts, and it should also be more sensitive than other performance measures to accounting changes in the short term. The balance sheet, on the other hand, would not be expected to change quickly in response to changes in accounting because shareholders' equity includes the effects of accounting choices made over the life of the firm and accounting changes are generally applied prospectively. Nevertheless, if changes in accounting have made measurement more comparable, then we would expect to see a narrowing of differences in balance sheet valuation over time.
Table 1 presents statistics on book-to-market ratios over the sample period. Following our earnings analysis, we regress mean-adjusted book-to-market ratios on the 14 country/ subperiod indicators. Again, we base our inference on whether coefficient estimates move closer to the mean in absolute value. Results (not tabulated) provide evidence of convergence, although it is not as strong as the evidence for earnings/price ratios. In the second period, cross-country differences were lower for five of the seven countries. Computing a rank sum statistic for the first half vs. the second half, the reduction in variance of medians is significant at the .05 level. Similarly, examining the relation between dispersion of book/ market ratios and time, there is a significant negative correlation over our sample period. Including GDP growth, sales growth, EPS growth, and real interest rates has no effect on the inference. Overall, there is some evidence of convergence in book-to-market ratios, but it is not as consistent as for earnings.
Our results suggest that earnings valuation across countries has become more similar over time. Results are robust to controls for underlying economic activity and are generally consistent across sample countries, but are strongest for Germany and Japan. The changes in multiples are robust to controls for the pricing of cash flows, suggesting that the effects are driven by the pricing of accruals, consistent with an increase in comparability of accounting over time. That conclusion is reinforced by evidence that the multiples applied to accruals have converged. However, significant predictable differences remain suggesting that, while comparability appears to have improved, the process is incomplete.
Results on the correlation between accruals and cash flows also indicate convergence, consistent with the conclusions based on the pricing multiples. Japan, Germany, and France show the most negative relation between cash flows and accruals, while the U.S. and Canada are the least negative. Again, evidence of a change is clearest for Japan and Germany.
While one cannot draw strong policy implications based on evidence of this type, our results raise the possibility that, even in the face of fairly haphazard global accounting standard development, accounting practices may be becoming more similar. Given the implicit and explicit pressures toward harmonization of financial reporting, it is perhaps not surprising that differences are declining. However, the results also suggest that the effects of traditional differences, while reduced, still persist.
The analysis is limited in many ways and is subject to numerous caveats. First, our conclusions are only as good as our methods. We apply approaches that have been used in the literature, attempt to control for as many variables as possible and consider a variety of analyses with similar conclusions. Nevertheless, we are unable to control for all potential omitted variables, so our results should be viewed with caution. Further, our sample is limited to the largest firms in the largest economies. Many of the more difficult regulatory issues relate to small young firms in emerging industries and less-developed markets where financial reporting is less developed and enforcement is limited.
TABLE 1 Descriptive Statistics for Sample Firms Number Median Mean 1st Qu. 3rd Qu. Panel A: Group 1 (Years 1987-1992) Earnings/Price Australia 657 .074 .096 .047 .106 Canada 1,601 .062 .071 .035 .091 Germany 682 .038 .042 .024 .057 France 580 .064 .080 .046 .096 United Kingdom 2,991 .076 .089 .052 .109 Japan 4,190 .017 .020 .010 .027 United States 11,713 .064 .073 .041 .092 Book/Market Australia 652 0.65 0.87 0.44 1.03 Canada 1,563 0.72 0.88 0.47 1.06 Germany 673 0.45 0.52 0.29 0.67 France 574 0.50 0.90 0.34 0.81 United Kingdom 2,953 0.56 0.75 0.35 0.95 Japan 4,149 0.27 0.30 0.17 0.40 United States 11,219 0.56 0.68 0.34 0.85 Total Assets ($ Millions) Australia 642 288 854 103 919 Canada 1,563 295 929 105 1,028 Germany 663 632 2,783 200 2,333 France 565 1,276 3,083 514 3,474 United Kingdom 2,925 217 925 82 755 Japan 4,105 1,008 3,078 412 2,742 United States 11,472 252 1,262 93 971 Return on Equity Australia 640 11.7 11.0 6.4 16.9 Canada 1,560 9.7 8.6 3.7 14.7 Germany 661 8.5 8.7 5.0 12.5 France 564 13.7 13.6 8.7 18.4 United Kingdom 2,917 15.2 16.3 7.4 22.8 Japan 4,083 6.8 7.3 4.6 9.4 United States 11,290 12.5 11.5 6.3 18.2 Number Median Mean 1st Qu. 3rd Qu. Panel B: Group 2 (Years 1994-1999) Earnings/Price Australia 897 .063 .072 .041 .087 Canada 1,777 .052 .066 .030 .082 Germany 1,339 .045 .068 .027 .076 France 1,485 .055 .066 .035 .085 United Kingdom 4,224 .063 .078 .042 .095 Japan 6,633 .031 .038 .017 .050 United States 11,564 .053 .061 .033 .077 Book/Market Australia 881 0.59 0.67 0.40 0.84 Canada 1,726 0.60 0.73 0.35 0.91 Germany 1,316 0.47 0.57 0.29 0.69 France 1,467 0.55 0.70 0.32 0.91 United Kingdom 4,065 0.44 0.59 0.24 0.78 Japan 6,560 0.76 0.91 0.51 1.16 United States 10,983 0.45 0.56 0.26 0.71 Total Assets ($ Millions) Australia 874 263 809 88 758 Canada 1,737 341 1,084 141 1,084 Germany 1,307 423 2,785 139 1,519 France 1,449 370 2,554 122 1,727 United Kingdom 4,132 194 1,037 71 726 Japan 6,497 769 2,553 355 2,091 United States 11,344 547 2,091 185 1,765 Return on Equity Australia 873 10.5 10.0 6.3 15.5 Canada 1,723 10.2 9.6 4.6 15.5 Germany 1,308 10.7 10.8 5.6 16.2 France 1,448 11.7 11.4 6.7 16.7 United Kingdom 4,126 15.1 15.7 7.8 24.4 Japan 6,504 3.6 3.1 1.5 6.0 United States 11,107 12.8 12.2 6.9 19.1 Data are from Global Vantage covering 1987-1999. Earnings/Price is earnings before extraordinary items divided by year-end market value and Book/Market is shareholders' equity divided by year-end market value, both with negative values excluded and extreme values truncated at 1 percent. Return on equity is earnings before extraordinary items divided by shareholders' equity. TABLE 2 Regression Results for Earnings to Price Ratios on Country/Subperiod Indicators Sub- Coeffi- Signi- Change Country period cient t-statistic ficance Significance Australia 1st .0284 14.13 .001 .001 2nd .0078 4.55 .001 Canada 1st .0034 2.65 .008 .417 2nd .0020 1.61 .107 Germany 1st -.0248 -12.55 .001 .001 2nd .0040 2.84 .005 France 1st .0129 6.03 .001 .001 2nd .0020 1.51 .131 U.K. 1st .0219 23.20 .001 .001 2nd .0134 16.83 .001 Japan 1st -.0476 -59.77 .001 .001 2nd -.0264 -41.69 .001 U.S. 1st .0058 12.09 .001 .001 2nd -.0028 -5.84 .001 Results are from the regression of mean-adjusted earnings/price ratios on 14 country/subperiod indicator variables, with no intercept. Earnings are before extraordinary items and price is as of year-end. The subperiods are 1987-1992 and 1994-1999, and there are 50,333 observations. Earnings/price ratios are adjusted by subtracting the subperiod mean of .067 in the first subperiod and .064 in the second. The change significance is from an F-statistic for the change in coefficient estimates between periods. TABLE 3 Regression Results for Cash Flows to Price Ratios on Country/Subperiod Indicators Sub- Coeffi- Signi- Change Country period cient t-statistic ficance Significance Australia 1st -.0213 -2.81 .005 .726 2nd -.0248 -3.99 .001 Canada 1st .0061 0.99 .325 .006 2nd -.0156 -3.13 .002 Germany 1st .0216 2.95 .003 .042 2nd .0397 7.95 .001 France 1st .0576 6.67 .001 .483 2nd .0506 10.05 .001 U.K. 1st .0169 4.57 .001 .001 2nd -.0151 -5.19 .001 Japan 1st -.0783 -22.67 .001 .001 2nd -.0130 -4.53 .001 U.S. 1st -.0025 -1.31 .190 .001 2nd -.0218 -12.21 .001 Results are from the regression of mean-adjusted cash flows/price ratios on 14 country/subperiod indicator variables, with no intercept. The subperiods are 1987-1992 and 1994-1999, and there are 50,487 observations. Cash flows from operations/price ratios are adjusted by subtracting the subperiod mean of 0.115 in the first subperiod and 0.113 in the second. The change significance is from an F-statistic for the change in coefficient estimates between periods. TABLE 4 Regression Results for Earnings/Price on Country/Subperiod Indicators with Controls for GDP Growth, Real Interest Rates, Earnings Growth, and Returns Sub- Coeffi- Signifi- Change Country period cient t-statistic cance Significance Australia 1st .0278 8.24 .001 .001 2nd .0102 3.63 .001 Canada 1st .0019 0.72 .474 .193 2nd .0042 1.62 .104 Germany 1st -.0213 -7.94 .001 .001 2nd -.0001 -0.03 .978 France 1st .0116 3.43 .001 .003 2nd .0030 1.25 .212 U.K. 1st .0195 12.33 .001 .001 2nd .0141 6.47 .001 Japan 1st -.0446 -26.70 .001 .001 2nd -.0319 -25.93 .001 U.S. 1st .0051 3.08 .002 .001 2nd .0004 0.24 .812 GDP Growth -.0016 -9.00 Int. Rate -.0002 -0.58 EPS Growth -.0028 -28.03 Return .0000 0.63 Results are from the regression of mean-adjusted earnings/price ratios on 14 country/subperiod indicator variables and controls for GDP growth, real interest rates, earnings growth, and returns, with no intercept. The subperiods are 1987-1992 and 1994-1999, and there are 42,300 total observations. Earnings/price ratios are adjusted by subtracting the subperiod mean of .074 in the first period and .069 in the second. The change significance is from an F-statistic for the change in coefficient estimates between periods. TABLE 5 Regression Results for Earnings/Price on Country/Subperiod Indicators with Controls for GDP Growth, Real Interest Rates, Earnings Growth, Returns, and Cash Flow/Price Sub- Coeffi- Signi- Change Country period cient t-statistic ficance Significance Australia 1st .0280 7.61 .001 .001 2nd .0104 3.37 .001 Canada 1st .0015 0.48 .629 .707 2nd .0024 0.79 .427 Germany 1st -.0223 -7.64 .001 .001 2nd -.0038 -1.58 .113 France 1st .0047 1.16 .245 .730 2nd .0035 1.28 .201 U.K. 1st .0228 13.08 .001 .001 2nd .0151 6.06 .001 Japan 1st -.0415 -19.71 .001 .001 2nd -.0284 -19.99 .001 U.S. 1st .0068 3.64 .001 .001 2nd .0009 0.41 .683 GDP Growth -.0014 -7.17 Int. Rate .0000 0.08 EPS Growth -.0031 -28.00 Return .0000 2.11 Results are from the regression of mean-adjusted earnings/price ratios on 14 country/subperiod indicator variables and controls for GDP growth, real interest rates, earnings growth, returns, and 14 controls for country and period cash flows from operations/price (not tabulated), with no intercept. The subperiods are 1987-1992 and 1994-1999, and there are 33,710 total observations. Earnings/price ratios are adjusted by subtracting the subperiod mean of .068 in the first subperiod and .066 in the second. The change significance is from an F-statistic for the change in coefficient estimates between periods. TABLE 6 Regression Results for Accruals/Price on Country/Subperiod Indicators with Controls for GDP Growth, Real Interest Rates, Earnings Growth, Returns, and Cash Flow/Price Sub- Coeffi- Signi- Change Country period cient t-statistic ficance Significance Australia 1st .0273 4.26 .001 .001 2nd .0011 0.21 .832 Canada 1st .0033 0.59 .553 .365 2nd -.0004 -0.07 .946 Germany 1st -.0189 -3.74 .001 .001 2nd -.0020 -0.47 .641 France 1st -.0006 -0.09 .925 .341 2nd .0051 1.09 .275 U.K. 1st .0296 9.76 .001 .001 2nd .0169 3.91 .001 Japan 1st -.0447 -12.23 .001 .001 2nd -.0212 -8.59 .001 U.S. 1st .0040 1.23 .217 .008 2nd .0004 0.10 .923 Results are from the regression of mean-adjusted total accruals/price ratios on 14 country/subperiod indicator variables and controls for GDP growth, real interest rates, earnings growth, returns, and 14 controls for country and period cash flows from operations/price, with no intercept. The subperiods are 1987-1992 and 1994-1999, and there are 33,914 total observations. Accruals/price ratios are adjusted by subtracting the subperiod mean of -.049 in both the first and second subperiods. The change significance is from an F-statistic for the change in coefficient estimates between periods. TABLE 7 Accrual/Cash Flow Correlations Country Subperiod Coefficient Z-Statistic Australia First -0.790 -4.92 Second -0.707 Canada First -0.655 -0.32 Second -0.594 Germany First -0.919 -3.52 Second -0.868 France First -0.897 -2.36 Second -0.886 U.K. First -0.795 -7.92 Second -0.686 Japan First -0.940 -11.73 Second -0.835 U.S. First -0.557 1.86 Second -0.565 Results are rank correlations of accruals and cash flows from operations, both deflated by total assets. The first subperiod is 1987-1992 and the second is 1994-1999. The Z-statistic is from a rank sum test of the significance of the change in correlations between periods.
We thank Katherine Schipper (editor), David A. Guenther (discussant), Allison Evans, Maria Nondorf, two anonymous reviewers, and workshop participants at The Accounting Review's Conference on Quality of Earnings, Emory University, University of Maryland, University of Michigan Spring Training, and The University of North Carolina at Chapel for helpful comments on earlier versions.
(1) David Cairns, former Secretary-General of the IASC, states, for example, "many European companies have recognized the need to make their financial reports more relevant to, and understandable by, international capital markets. They have responded to this need by going well beyond the limited requirements of the national accounting laws and standards of the time" (Calms 1997). Choi et al. (1999) make a similar case.
(2) Cairns (1997) notes, for example, "when we look at the way countries or companies account for particular transactions or events, it is increasingly difficult to distinguish in a systematic way so-called Anglo-American accounting from Continental European accounting or American accounting from, say, German accounting." Similarly, Choi et al. (1999) indicate, "harmonization debates aside, all dimensions of accounting are being harmonized worldwide" (emphasis in the original).
(3) While these goals have some intuitive appeal, we do not claim that they are necessarily optimal or that our constructs capture them perfectly. Further, our approach focuses on measurement issues and their effects on bottom-line net income. Supplemental disclosure and other financial statement issues are also important, but are even more difficult to assess and have not (at least to date) been the subject of as much concern as income measurement issues.
(4) Some studies (e.g., Alford et al. 1993) use data for a few years earlier, but licensing agreements and transfers of ownership between data providers limit our ability to access those data. We compared Global Vantage, Datastream, and Worldscope, but data availability (especially for early years) is best with Global Vantage. We avoided mixing data providers to increase the likelihood that the data are prepared on a consistent basis.
(5) Observations in each country are generally spread across a range of industries and, because the comparison is over time, cross-country differences in industry concentration should not affect results. Including industry controls does not affect our conclusions.
(6) Results are very similar if loss observations are included. In addition, results are robust to analysis based on medians and rank regression, suggesting that remaining large observations do not drive results.
(7) France is more difficult to categorize. Based on simulations of European accounting practices, Simmonds and Azieres (1989) find that Germany is consistently more conservative than the U.K., but France generally lies between Germany and the U.K., and in some cases is less conservative than the U.K. (see also Alexander and Archer 1991). Ball, Kothari, and Robin (2000) also find that France lies between Germany and the U.K. in terms of earnings valuation.
(8) We use groupings to increase the power of the tests. As we discuss later, results are robust to alternate groupings and are consistent on a year-by-year basis, suggesting that a subset of years does not drive the results. We use a regression specification for consistency with later analysis when we include controls for other potential factors. Results are similar for a standard t-test comparison of means.
(9) This approach implicitly assumes that the relevant benchmark for assessing convergence is whether countries have become more similar. An alternate approach would be to assess whether accounting worldwide has moved toward a benchmark set of practices, like that represented by U.S. GAAP. Redefining convergence as movement toward U.S. multiples yields very similar results because the U.S. is close to the average and stable over the sample period.
(10) An issue with the statistical inference is potential autocorrelation of residuals. Residual autocorrelation is approximately 0.3 and applying a Cochrane-Orcutt adjustment does not affect the inference.
(11) See, for example, Choi et al. (1999) for a discussion. For a discussion of progress toward consistent standards for specific countries during the sample period see Curran (1996) on Australia; CA Magazine (1996) on Canada; Doupnik (1992) and Pape (1997) on Germany; Accountancy (1997a) on France; Shibata (1998) on Japan; and Accountancy (1997b) on the U.K.
(12) The coefficient of variation is the standard deviation of the seven-country mean earnings/price ratios divided by the grand mean of the seven-country means to control for movements in average earnings/price ratios over time. Results are not sensitive to dividing by the mean.
(13) The calculation is net income before extraordinary items plus depreciation minus the change in noncash assets (other than fixed assets) plus the change in liabilities (other than long-term debt). The decreased number of observations reflects missing data. To mitigate the effect of extreme observations, accruals in excess of market value in absolute value were deleted (these are generally large negative accruals consistent with "big baths").
(14) We were unable to obtain growth forecasts for most of our sample firms, especially in the first period, and use realized rates instead. Including returns in the regression should partially capture the effects of differences between forecasted and realized growth. In addition, we replicated the analysis using year-ahead and two-year-ahead realized earnings growth. While that approach limits our sample period by eliminating the later years, results are very similar.
(15) Another possibility is that changes in earnings multiples are related to changes in the market multiple applied to the book value of equity. Results are robust to including controls for changes in the multiple on book value.
(16) Specifically, we include 14 control variables, one for each country/subperiod, which take on a value of the cash flow/price ratio if the observation is for that country in that subperiod, and 0 otherwise. This allows the coefficient to vary between countries and between subperiods.
(17) The pattern is similar for the standard deviation plotted over time, but deflating by the mean ensures that the pattern is not driven by changes in average multiples over time. Results are consistent for a nonparametric approach using the interquartile range divided by the median. A similar analysis based on earnings/price ratios indicates reductions in dispersion in the later portion of the sample period (although less pronounced) as well as in the earlier period.
(18) In an earlier draft, we also included an analysis based on the proportion of small positive earnings as in Burgstahler and Dichev (1997). That approach is more difficult to interpret since it assumes that firms in all countries are managing toward the same objective, but conclusions were generally consistent with those presented here.
(19) We chose total assets as a deflator rather than market value so that our measure is unaffected by stock price movements. Results are consistent deflating by market value.
(20) This is not to suggest that U.S. firms necessarily manage earnings less than those in other countries since, as argued in Brown and Higgins (2001), they may be managing earnings to meet analyst forecasts.
(21) These results are generally consistent with those in Ball, Kothari, and Robin (2000), which focuses on differences in conservatism (which they interpret as evidence of smoothing) across countries. In supplemental analysis, they find evidence of reduced conservatism for Australia, Germany, France, and U.K., but not for the U.S., Canada, or Japan. Bhattacharya et al. (2001) also compare earnings management over time using different measures and provide evidence that it is generally declining.
(22) We also examined changes in the explanatory power of earnings for returns over time, but found no clear pattern across the two subperiods. However, we were unable to document evidence of systematic cross-country differences in either subperiod, suggesting that this approach may lack power.
Accountancy. 1997a. French reforms going through. (February): 12.
--. 1997b. U.K. businesses feel IASC pressure. (March): 8.
Alexander, D., and S. Archer. 1991. European Accounting Guide. San Diego, CA: HBJ Professional Publishing.
Alford, A., J. Jones, R. Leftwich, and M. Zmijewski. 1993. The relative information content of accounting disclosures in different countries. Journal of Accounting Research 31 (Supplement): 183-223.
Ball, R., S. P. Kothari, and A. Robin. 2000. The effect of international institutional factors on properties of accounting earnings. Journal of Accounting and Economics 29 (October): 1-51.
Ball, R., A. Robin, and J. S. Wu. 2000. Incentives versus standards: Properties of accounting income in four East Asian countries, and implications for acceptance of IAS. Working paper, University of Rochester.
Barth, M., and G. Clinch. 1996. International accounting differences and their relation to share prices: Evidence from U.K., Australian, and Canadian firms. Contemporary Accounting Research 13 (Spring): 135-170.
Beatty, A., B. Ke, and K. Petroni. 2002. Earnings management to avoid earnings declines across publicly and privately held banks. The Accounting Review 77 (3): 547-570.
Bhattacharya, U., H. Daouk, and M. Welker. 2001. The world price of earnings opacity. Working paper, Indiana University.
Biener, H. 1994. What is the future of mutual recognition of financial statements and is comparability really necessary? European Accounting Review 3: 335-342.
Brown, L., and H. Higgins. 2001. Managing earnings surprises in the U.S. versus 12 other countries. Journal of Accounting and Public Policy 20 (Winter): 373-398.
Burgstahler, D., and I. Dichev. 1997. Earnings management to avoid earnings decreases and losses Journal of Accounting and Economics 24 (December): 99-126.
CA Magazine. 1996. Champion of Canadian standards. (January/February): 10.
Cairns, D. 1997. The future shape of harmonization: A reply. European Accounting Review 6 (2): 305-348.
Choi, F. D. S., C. Frost, and G. Meek. 1999. International Accounting. 3rd edition. Upper Saddle River, NJ: Prentice Hall.
Cochrane, J. 1994. Are U.S. regulatory requirements for foreign firms appropriate? Fordham International Law Journal 17: 58-67.
--, J. Shapiro, and J. Tobin. 1996. Foreign equities and U.S. investors: Breaking down the barriers separating supply and demand. Stanford Journal of Law, Business and Finance (Summer): 241263.
Curran, B. 1996. An Australian comparison. Accountancy (September): 64-65.
Dhaliwal, D., D. Guenther, and M. Trombley. 1999. Inventory accounting methods and earnings-price ratios. Contemporary Accounting Research (Fall): 419-436.
Doupnik, T. 1992. Recent innovations in German accounting practice through the integration of EC directives. Advances in International Accounting 5: 75-103.
Economist. 1999. Rupert laid bare. (March 20): 63-64.
Financial Accounting Standards Board (FASB). 1996. The IASC-U.S. Comparison Project: A Report on the Similarities and Differences between IASC Standards and U.S. GAAP. Edited by C. Bloomer. Stamford, CT: FASB.
French, K. R., and J. M. Poterba. 1991. Were Japanese stock prices too high? Journal of Financial Economics 29 (October): 337-362.
Frost, C., and K. Ramin. 1997. Corporate financial disclosure: A global assessment. In International Accounting and Finance Handbook, edited by F. D. S. Choi. New York, NY: John Wiley & Sons.
Guenther, D., and D. Young. 2000. The association between financial accounting measures and real economic activity: A multinational study. Journal of Accounting and Economics 29 (February): 53-72.
Jenkins, E. 1999. Financial reporting in a global capital market world. Financial Accounting Series 198-A (June 29): 2-6.
Joos, P., and M. Lang. 1994. The effects of accounting diversity: Evidence from the European Union. Journal of Accounting Research 32 (Supplement): 141-175.
Lang, M., J. Raedy, and M. Yetman. 2002. How representative are cross-listed firms? An analysis of firm performance and accounting quality. Working paper, The University of North Carolina at Chapel Hill.
Leuz, C., D. J. Nanda, and P. Wysocki. 2001. Investor protection and earnings management: An international comparison. Working paper, University of Pennsylvania.
Myers, L., and D. Skinner. 2001. Earnings momentum and earnings management. Working paper, University of Michigan.
Pape, J. 1997. A milestone in German accounting. Accountancy 119 (May): 58-59.
Pownall, G., and K. Schipper. 1999. Implications of accounting research for the SEC's consideration of International Accounting Standards for U.S. securities offerings. Accounting Horizons 13 (September): 259-280.
Shibata, Y. 1998. Turnabout in Japan. Global Finance 12 (November): 44-45.
Simmonds, A., and O. Azieres. 1989. Harmonisatie van de verslaggeving in Europa: Voltooid in het jaar 2000. London, U.K.: Touche Ross Europe.
Speidell, L. S., and V. B. Bavishi. 1992. GAAP arbitrage: Valuation opportunities in International Accounting Standards. Financial Analysts Journal (November/December): 58-66.
Judy Land Mark H. Lang The University of North Carolina at Chapel Hill
|Printer friendly Cite/link Email Feedback|
|Author:||Land, Judy; Lang, Mark H.|
|Date:||Dec 1, 2002|
|Previous Article:||Discussion of the relation between auditors' fees for nonaudit services and earnings management.|
|Next Article:||Discussion of empirical evidence on the evolution of international earnings.|