Analyst Following and Forecast Accuracy After Eliminating the 20-F Reconciliation Between IFRS and U.S. GAAP

Abstract

In 2007, the U.S. Securities and Exchange Commission (SEC) decided to allow foreign private issuers to file financial statements prepared according to International Financial Reporting Standards (IFRS) without reconciliation to U.S. Generally Accepted Accounting Principles (GAAP). Using a sample of foreign private issuers from 35 countries/regions during the period of 2005 to 2008, this article investigates how the elimination of the 20-F reconciliation affects financial analysts. We find that it significantly reduces analyst coverage but has no impact on forecast accuracy. We show that analysts who are greatly affected are more likely to terminate their coverage of IFRS firms after the SEC’s rule than other analysts. In addition, we hypothesize and find that eliminating the 20-F reconciliation has a greater impact on firms whose 20-F reconciliation is more useful to analysts. For these firms, the elimination of the 20-F reconciliation significantly reduces both analyst coverage and forecast accuracy. Overall, our results suggest that the elimination of the 20-F reconciliation imposes costs on financial analysts.

Keywords

financial analysts International Financial Reporting Standards (IFRS)U.S. GAAP Form 20-F reconciliation

Introduction

Prior to 2007, all foreign private issuers (FPIs) listed in the U.S. stock exchanges were required by the Securities and Exchange Commission (SEC) to reconcile to U.S. Generally Accepted Accounting Principles (GAAP) in Form 20-F, if their financial statements are prepared under non-U.S. accounting standards. The reconciliation to U.S. GAAP is commonly referred to as the 20-F reconciliation. The requirement for the reconciliation was initiated in 1982 to meet U.S. investors’ needs for information and for opportunities to invest in foreign securities (SEC, 2005). In November 2007, the U.S. SEC ruled to eliminate the requirement for the FPI to provide reconciliation to U.S. GAAP if the issuer prepares its financial statements according to International Financial Reporting Standards (IFRS) as adopted by the International Accounting Standards Board (IASB).

This study investigates how the elimination of the 20-F reconciliation affects financial analysts. The 20-F reconciliation bridges the gap between non-U.S. accounting rules and U.S. GAAP, and its elimination presumably affects analyst behavior. Prior studies show unambiguously that differences in accounting rules impose costs on analysts (Ashbaugh & Pincus, 2001; Bae et al., 2008; Guan et al., 2006; Tan et al., 2011). When the 20-F reconciliation becomes unavailable, it is difficult or even impossible for financial analysts to compare the performance of foreign firms with that of U.S. firms on an apples-to-apples basis (CFA Institute, 2007). Therefore, it negatively affects their forecast accuracy and discourages analysts from continuing covering the firm. In addition, the elimination may reduce investors’ demands for these stocks (Kim et al., 2012). As analysts are incentivized to cater to the information needs of investors, when investors’ demands for these stocks are reduced, analysts have lower incentives to continue following them. We therefore expect that the elimination would lead to a drop in both analyst coverage and forecast accuracy.

The elimination of the 20-F reconciliation has differential impacts on analysts. Conceivably, the elimination is more costly for analysts who find the information in the 20-F reconciliation useful than for analysts who do not. We therefore predict that the likelihood of terminating coverage of IFRS firms (firms that use IFRS to prepare their financial statements) after the elimination of the 20-F reconciliation is higher for analysts who find the reconciliation useful. We infer the usefulness of the reconciliation to financial analysts through two proxies. First, we use analysts’ response to the release of the 20-F reconciliation as an indicator of how analysts perceive the reconciliation. If analysts revise their forecasts around the 20-F filing date (the time when the 20-F reconciliation first becomes publicly available¹), it is a strong signal that the reconciliation is regarded as useful. We also infer the usefulness of the reconciliation from analysts’ familiarity with the accounting rules through their coverage. If they cover both IFRS firms and U.S. GAAP firms (firms that use U.S. GAAP to prepare their financial statements), they are likely familiar with both sets of rules and the elimination of the 20-F reconciliation is less costly to them. We hypothesize that analysts who find the reconciliation useful are more likely to terminate their coverage of IFRS firms than analysts who do not.

We further examine how the impact of eliminating the 20-F reconciliation on financial analysts is influenced by characteristics of the 20-F reconciliation. Our hypothesis is that eliminating the 20-F reconciliation results in a more pronounced drop in analyst following and forecast accuracy for firms with more informative 20-F reconciliation than for firms with less informative 20-F reconciliation. As eliminating more informative 20-F reconciliation imposes greater costs on analysts and reduces demands of information from investors to a greater extent, we expect to observe a more pronounced decrease in analyst following and forecast accuracy.

Our empirical analysis is based on a sample of 492 firm-year observations (1,726 firm-year-analyst observations) pertaining to 123 distinct U.S.-listed foreign firms from 35 countries/regions in the period of 2005 to 2008. We identify treatment firms as firms that use IFRS as adopted by IASB and stop providing the reconciliation after 2007, whereas control firms are firms that use either U.S. GAAP or foreign domestic GAAP and experience no change in their provision of the 20-F reconciliation. We compare changes in analyst coverage and forecast accuracy after the elimination of the 20-F reconciliation between treatment firms and control firms.

Our results can be summarized as follows. We find that eliminating the 20-F reconciliation has no statistically significant impact on forecast accuracy.² However, it significantly reduces the number of analysts following the firm. Our results are consistent with our assertion that the analysis based on forecast accuracy alone is biased toward insignificant findings.

To test our prediction that the likelihood of terminating coverage of IFRS firms is higher for analysts who revise their forecasts around the 20-F filing date and is lower for analysts who cover both IFRS firms and U.S. GAAP firms, we run a multivariate logistic regression. After controlling for various firm and analyst characteristics, we find that the odds of terminating coverage of the IFRS firm after the elimination of the 20-F reconciliation are more than four times higher for analysts who revise their forecasts around the 20-F filing date than for analysts who do not. In addition, the odds are 88% lower for analysts who cover both IFRS firms and U.S. GAAP firms than those for other analysts who cover only IFRS firms. Our results suggest that analysts who are affected to a greater extent by the elimination of the 20-F reconciliation are more likely to terminate their coverage.

We next test the hypothesis that eliminating the 20-F reconciliation results in a more pronounced drop in analyst following and forecast accuracy for firms with more informative 20-F reconciliations. Results from all four measures of the informativeness of the 20-F reconciliation support this hypothesis. These results highlight the importance of analyzing subsamples of firms for a comprehensive assessment of the impact of eliminating the 20-F reconciliation. Collectively, our results suggest that the elimination of the reconciliation imposes costs on financial analysts.

In addition to our main tests, we examine the impact of eliminating 20-F reconciliation on analyst forecast dispersion and conduct several sensitivity tests. Our results are available in Supplemental Appendix.

Our article is directly related to prior studies exploring the consequences of the SEC’s rule to eliminate the 20-F reconciliation requirement for IFRS firms (Byard et al., 2010; Chen, Deng, et al., 2015; Chen & Khurana, 2015; Hansen et al., 2010; Kang et al., 2012; Kim et al., 2012; Lin & Huang, 2014; Yang & Henry, 2013; Yu, 2011). Although these studies focus on its impact on various aspects such as information environment, earnings attributes, voluntary disclosure, and shareholder wealth, we focus exclusively on its impact on financial analysts. Financial analysts are the subject of considerable academic interest and are widely regarded as sophisticated and influential capital market participants.³ Our study thus examines the consequence of the SEC’s rule from an important and relevant perspective. Furthermore, we identify an individual analyst’s response to the 20-F reconciliation and condition our analysis on it, whereas prior studies examine the impact of the SEC’s rule on an aggregate market level. Our perspective allows us to provide a potentially finer description of the impact of the rule.

Our article is indirectly related to prior studies examining investors’ response to the 20-F reconciliation (Chen & Sami, 2008, 2010; Gordon et al., 2009; Harris & Muller, 1999; Henry et al., 2009; Jiang et al., 2010; Plumlee & Plumlee, 2007). Although results from the value-relevance framework generally suggest the usefulness of the 20-F reconciliation, results from event studies are mixed. We contribute to this line of literature by showing that eliminating the 20-F reconciliation imposes costs on financial analysts.

Our article also contributes to the literature on financial analysts. A long line of literature examines the characteristics of analysts’ forecasts, analysts’ incentives, and consequences of their behaviors (see Beyer et al., 2010, for a review). Recent studies have investigated how analysts respond to regulatory changes and differences in accounting rules (e.g., Ashbaugh & Pincus, 2001; Bae et al., 2008; Byard et al., 2011; Tan et al., 2011). Our article extends this line of literature by providing evidence on how analysts respond to the elimination of the 20-F reconciliation. Our results suggest that it results in lower analyst coverage in general and lower forecast accuracy for firms whose 20-F reconciliation is more useful.

The rest of the article proceeds as follows. Section “Hypothesis Development” develops hypotheses. Section “Sample Formation and Research Design” covers sample formation and research design. Section “Empirical Results” discusses empirical results. Section “Conclusion” concludes.

Hypothesis Development

Our study empirically examines how the elimination of the 20-F reconciliation affects financial analysts. There are two important reasons why the elimination has an impact on financial analysts. First, it imposes costs on financial analysts and therefore reduces analyst coverage and forecast accuracy. The elimination of the 20-F reconciliation renders it more difficult for financial analysts to compare the performance of foreign firms with that of U.S. firms.⁴ Given the importance of U.S. economy in the world, for many industries, U.S. firms serve as an appropriate benchmark. Comparison with U.S. firms in the same industry disentangles the firm-specific factor from the industry-wide factor, helping financial analysts to gauge the foreign firm’s current performance and generate insights useful for predicting its future performance.⁵ The importance of financial statement comparability is documented in De Franco et al. (2011), who find that greater comparability lowers analysts’ information acquisition costs, leading to higher analyst following, more accurate forecasts, and lower forecast dispersion. When the 20-F reconciliation becomes unavailable, such a comparison between U.S. firms and foreign firms becomes more difficult or even impossible for financial analysts. It negatively affects their forecast accuracy and discourages analysts from continuing covering the firm.

Second, the elimination may reduce demands for the information of foreign firms whose 20-F reconciliation becomes unavailable after the SEC’s rule. As argued in Kim et al. (2012), eliminating the reconciliation may reduce the comparability of financial statements issued by foreign cross-listed and U.S. domestic companies. This may discourage U.S. investors from trading IFRS firms’ shares. According to Chen and Sami (2008, 2010), U.S. investors find earnings reconciliation from IAS to U.S. GAAP informative in their trading decisions. As they rely on the reconciliation in their trading, the absence of the reconciliation reduces investors’ incentives to trade these stocks and their demands for analyst research on these stocks. Consequently, analysts have diminished incentives to follow these stocks. Based on the two arguments, we expect that the elimination will lead to a drop in both analyst coverage and forecast accuracy.

Our first hypothesis is thus as follows:

Hypothesis 1 (H1): Analyst coverage and forecast accuracy drop after the elimination of the 20-F reconciliation.

Our hypothesis is not without tension. Firms may change their disclosure policies after the elimination of the 20-F reconciliation. For example, Yu (2011) documents an increase in voluntary disclosure after the elimination. It is therefore possible that the loss of the 20-F reconciliation is offset by the increased voluntary disclosure, leading to no change in analyst behavior. In sum, it is an empirical issue whether eliminating the 20-F reconciliation affects analyst following and forecast accuracy.

The impact of eliminating the 20-F reconciliation is likely to differ across analysts. The elimination may result in significant information loss for some analysts, for example, analysts who do not understand the differences between IFRS and U.S. GAAP, whereas it may have a negligible impact on other analysts, for example, analysts who are well informed of the accounting differences. Analysts who are greatly affected by the elimination anticipate a significant drop in forecast accuracy and a substantial increase in the cost of providing coverage. They are likely to terminate their coverage and the expected drop in forecast accuracy becomes effectively unobservable. In contrast, analysts who are not affected by the elimination anticipate no change in forecast accuracy and the cost of providing coverage. They are likely to continue their coverage and, as expected, exhibit no change in their forecasting performance. Thus, results based on the properties of observable analyst forecasts are biased toward concluding that the elimination of the 20-F reconciliation has no impact.

Our discussions above lead to the second hypothesis expressed as follows:⁶

Hypothesis 2 (H2): The likelihood of terminating coverage of IFRS firms is higher for analysts who find the 20-F reconciliation useful.

We next hypothesize that eliminating the 20-F reconciliation results in a more pronounced drop in analyst following and forecast accuracy for firms with more informative 20-F reconciliations than for firms with less informative 20-F reconciliations. Intuitively, eliminating more informative 20-F reconciliations imposes greater costs on analysts and reduces investors’ demands for information to a greater extent, leading to a more pronounced decrease in analyst following and forecast accuracy.

Our third hypothesis can be expressed as follows:

Hypothesis 3 (H3): Eliminating the 20-F reconciliation results in a more pronounced drop in analyst following and forecast accuracy for firms whose 20-F reconciliation is more informative.

Sample Formation and Research Design

Sample Formation

We form our initial sample of U.S.-listed foreign firms by merging the Compustat Global database, the Compustat North America database, CRSP, and IBES. We require that the firm list on at least one of the three major stock exchanges in the U.S. (New York Stock Exchange [NYSE], American Stock Exchange [AMEX], and NASDAQ) be covered on the IBES detailed file and have data available between 2005 and 2008.⁷

We then search the SEC’s EDGAR database for the 20-F filings. All of our sample firms file Form 20-F with the SEC and report in the 20-F form the accounting standard (U.S. GAAP/domestic GAAP/IFRS) used to prepare financial statements. We identify accounting standards through firms’ 20-F filings. We eliminate firms that switch their accounting standards during our sampling period because switches in accounting standards may affect financial analysts and therefore contaminate our results.

For each firm-year observation, we retrieve all 1-year-ahead earnings forecasts issued during the 6-month period prior to the fiscal year end. This period is specified to assure that the 20-F filings of the prior year are available to analysts when they issue forecasts, because the SEC requires that Form 20-F be filed within 6 months of the fiscal year end. If an analyst issues more than one forecast during this period, we retain only the last forecast. We drop all observations whose analysts or brokerage firms disappear from the IBES database after 2007 to alleviate the concern that our analysis is picking up the effect of laid-off analysts or the shrinkage in the financial services industry during our sampling period.

To study the impact of eliminating the 20-F reconciliation on analysts, we identify firms that use IFRS as treatment firms, and firms that use other accounting rules (e.g., domestic GAAP) as control firms. The SEC’s rule does not affect control firms, but it eliminates the requirement of the 20-F reconciliation for treatment firms. Consistent with Kim et al. (2012), in our sample, none of our treatment firms provide 20-F reconciliation after the SEC’s rule becomes effective.⁸ For the firm-year analysis, we require the following variables to be nonmissing for all firm-year observations: Ana_follow, LOGMV, BM, Lev, STD_ROA, Sec_issue, Stock_ret, UE, CHROA, EARNSKEW, and RESTRUCT. We require the same firm be included in the sample for the entire sampling period. For the firm-year-analyst analysis, we additionally require the following variables to be nonmissing: ACCURACY, Horizon, Firm_exp, Gen_exp, Nfirm, and US (please refer to Supplemental Appendix for detailed variable definitions). Our final sample consists of 492 firm-year observations, of which 160 observations pertain to 40 treatment firms and 332 pertain to 83 control firms. All continuous variables are winsorized at the top and bottom one percentile.

Research Design

Models (1) and (2) are used to investigate the effect of eliminating the 20-F reconciliation on analyst following and forecast accuracy, respectively. They are specified as follows:

\begin{matrix} \log {(Ana_follow)}_{j, t} = α_{0} + α_{1} TREA T_{j, t} + α_{2} TREA T_{j, t} \times AFTE R_{j, t} + α_{3} AFTE R_{j, t} + α_{4} LOGM V_{j, t} \\ + α_{5} B M_{j, t} + α_{6} Le v_{j, t} + α_{7} STD_RO A_{j, t} + α_{8} Sec_issu e_{j, t} + α_{9} Stock_re t_{j, t} \\ + α_{10} U E_{j, t} + α_{11} CHRO A_{j, t} + α_{12} EARNSKE W_{j, t} + α_{13} RESTRUC T_{j, t} \\ + Industry dummies + Year dummies + ε_{j, t} \end{matrix}

(1)

\begin{matrix} ACCURAC Y_{i, j, t} = α_{0} + α_{1} TREA T_{j, t} + α_{2} TREA T_{j, t} \times AFTE R_{j, t} + α_{3} AFTE R_{j, t} + α_{4} LOGM V_{j, t} \\ + α_{5} B M_{j, t} + α_{6} Le v_{j, t} + α_{7} STD_RO A_{j, t} + α_{8} Sec_issu e_{j, t} + α_{9} Stock_re t_{j, t} \\ + α_{10} U E_{j, t} + α_{11} CHRO A_{j, t} + α_{12} EARNSKE W_{j, t} + α_{13} RESTRUC T_{j, t} \\ + α_{14} \log (Ana_follow)_{j, t} + et_α_{15} Horizo n_{i, j, t} + α_{16} \log (Firm_\exp)_{i, j, t} \\ + α_{17} \log (Gen_\exp)_{i, t} + α_{18} \log {(Nfirm)}_{i, t} + α_{19} U S_{i, t} \\ + Industry dummies + Year dummies + ε_{i, j, t} \end{matrix}

(2)

where Ana_follow is one plus the number of distinct analysts issuing 1-year-ahead earnings forecasts during the 6 months prior to the fiscal year end.^9,10TREAT equals 1 for firms that use IFRS as adopted by IASB during the sampling period and stop providing the reconciliation after the SEC’s rule, and 0 otherwise. AFTER equals 1 for fiscal years ending after December 15, 2007 (the effective date of the SEC’s rule eliminating the reconciliation requirement for IFRS firms), and 0 otherwise. LOGMV is the log value of the market capitalization in millions, computed as stock price (Compustat item PRCC_F) times total shares outstanding at the fiscal year end (Compustat item CSHO). BM is the ratio of the book value of equity (Compustat item CEQ) to the market value of equity (stock price [Compustat item PRCC_F] times total shares outstanding at the fiscal year end [Compustat item CSHO]) at the fiscal year end. Lev is the sum of long-term debt (Compustat item DLTT) and debt in current liabilities (Compustat item DLC) deflated by total assets (Compustat item AT) at the fiscal year end. STD_ROA is the standard deviation of ROA (income before extraordinary items [Compustat item IB] divided by the beginning-of-year total assets [Compustat item AT]) over the prior 5 years. UE is unexpected earnings, calculated as the absolute value of the difference between the current year’s earnings per share and the last year’s earnings per share, divided by the price at the beginning of the fiscal year. CHROA refers to the change in ROA, calculated as the difference between the current year’s ROA and the last year’s ROA. EARNSKEW is the earnings skewness based on the time series of earnings per share over the past 10 years, using the statistical definition of skewness. RESTRUCT is a dummy variable that equals 1 if a firm reported a restructuring activity during the year, and 0 otherwise. Sec_issue equals 1 if the firm issues equity or debt greater than 5% of total assets during the year, and 0 otherwise. Stock_ret is the annual stock return, adjusted for contemporaneous annual market return. ACCURACY reflects the forecast accuracy, a higher value of which indicates a more accurate forecast. It is computed as minus one hundred multiplied by the absolute value of the difference between the analyst forecast and the actual earnings of the firm scaled by the stock price at the end of the prior fiscal year. The actual earnings are as reported by the IBES database. Minus 100 is used to facilitate reporting and exposition. Horizon reflects the forecast horizon. It is computed as the time interval in years between the analyst’s forecast date and the fiscal year end date. Firm_exp represents the firm-specific experience, computed as the time interval in years between the analyst’s first forecast and the current forecast for the firm. Gen_exp represents the general experience, computed as the time interval in years between the analyst’s first forecast for any firm and the current forecast for the firm. Nfirm indicates the number of firms followed by an analyst. It is computed as the number of distinct firms for which the analyst issues at least one 1-year-ahead earnings forecast in the 6 months prior to the fiscal year end. US indicates the location of the analyst. It equals 1 for analysts located in the United States, and 0 otherwise. Analyst location data are obtained from Nelson’s Directory of Investment Research and matched with the I/B/E/S database in the same way as Bae et al. (2008) and Tan et al. (2011). The subscripts i, j, and t stand for analyst i, firm j, and year t, respectively.

Model (1) controls for the following firm characteristics: the market value, the book-to-market ratio, the leverage ratio, the standard deviation of ROA, the security issuance dummy, stock return, unexpected earnings, change in ROA, earnings skewness, and restructuring activity.¹¹ These firm characteristics affect analyst following, according to the literature (e.g., Barth et al., 2001; Bhushan, 1989; Chaney et al., 1999; Chen, Krishnan & Sami, 2015; Duru & Reeb, 2002; Lang et al., 2004; Lang & Lundholm, 1996; O’Brien & Bhushan, 1990). Model (2) additionally controls for analyst characteristics: the number of analysts following the firm, forecast horizon, firm-specific experience, general experience, the number of firms followed by the analyst, and analyst location. Prior studies (e.g., Bae et al., 2008; Brown, 2001; Clement, 1999; Clement & Tse, 2005; Jacob et al., 1999; Mikhail et al., 1997; Tan et al., 2011) show that these analyst characteristics affect individual analyst forecast accuracy. Our inferences are based upon standard errors clustered at firm level to account for residual dependence.

Our model specification reflects the difference-in-difference research design. In both models, our focus is on α₂, the coefficient of the interaction term, TREAT×AFTER. It measures the difference in the post-elimination change between treatment firms and control firms. If H1 is true, we expect the coefficient to be negative and significant in both models.

Empirical Results

Descriptive Statistics

Table 1 reports the sample selection procedure and descriptive statistics for our sample. Panel A summarizes our sample selection procedure. We start with 1,328 firm-year observations of FPIs covered by the Compustat Global database, the Compustat North America database, CRSP, and the IBES database from 2005 to 2008. We delete 508 observations with missing variables calculated from Compustat, CRSP, and IBES. We further eliminated 200 firm-year observations with either missing data in 20-F filings or firms that switch accounting standard during our sampling period. We get rid of 128 observations for firms not included for the entire sampling period. Our final sample consists of 492 firm-year observations.

Table 1.

Descriptive Statistics.

Panel A: Sample Selection.
Sample selection process						No. of firm-year observations
U.S.-listed foreign firms covered by the Compustat Global database, the Compustat North America database, CRSP, and IBES from 2005 to 2008						1,328
Less: firm-year observations with missing data in Compustat						(41)
Less: firm-year observations with missing data in CRSP						(113)
Less: firm-year observations with missing data in IBES						(354)
Less: firm-year observations with missing data in 20-F filings, or firms that switch their accounting standards during our sampling period						(200)
Less: firm-year observations of firms NOT included for the entire sampling period						(128)
Final sample						492
Panel B: Summary Statistics.
Variables	Obs.	M	SD	Q1	Median	Q3
Ana_follow	492	5.352	5.052	2	4	7
TREAT (dummy)	492	0.325	0.469	0	0	1
AFTER(dummy)	492	0.467	0.499	0	0	1
LOGMV	492	8.862	1.890	7.589	9.038	10.293
BM	492	0.573	0.439	0.277	0.459	0.710
Lev	492	0.231	0.172	0.094	0.222	0.349
STD_ROA	492	0.051	0.051	0.018	0.034	0.067
Sec_issue	492	0.472	0.500	0	0	1
Stock_ret	492	0.102	0.422	−0.130	0.048	0.240
UE	492	0.042	0.066	0.010	0.022	0.043
CHROA	492	−0.011	0.062	−0.037	−0.004	0.017
EARNSKEW	492	0.008	1.146	−0.666	0.164	0.761
RESTRUCT	492	0.242	0.429	0	0	0
ACCURACY	1,726	−0.838	1.489	−0.852	−0.311	−0.098
Horizon	1,726	0.131	0.110	0.083	0.083	0.167
Firm_exp	1,726	3.520	3.389	1.000	2.667	5.167
Gen_exp	1,726	10.437	7.422	4.417	8.083	15.833
Nfirm	1,726	11.760	8.287	5	11	17
US	1,726	0.688	0.464	0	1	1

Note. Panel A summarizes the sample selection procedure. Panel B presents the summary statistics. For Ana_follow, TREAT, AFTER, LOGMV, BM, Lev, STD_ROA, Sec_issue, Stock_ret, UE, CHROA, EARNSKEW, and RESTRUCT, the descriptive statistics are based on 492 firm-year observations. For the remaining variables, the descriptive statistics are based on 1,726 firm-year-analyst observations. All the continuous variables are winsorized at the top and bottom one percentile. Please refer to Supplemental Appendix for detailed variable definitions.

Panel B reports the mean, standard deviation, first quartile, median, and third quartile for both firm-level and analyst-level variables. In untabulated tests, we compare treatment firms and control firms. Consistent with Kim et al. (2012), treatment firms are different from control firms in certain firm characteristics, such as the market value, book-to-market ratio, and leverage. They also differ in the likelihood of reporting restructuring activities. These results suggest a need to control for these variables in our regressions.

Testing H1

H1 posits that analyst coverage and forecast accuracy drop after the elimination of the 20-F reconciliation. To test it, we run the regressions specified in Models (1) and (2) in Section “Research Design” and report our results in Table 2.

Table 2.

The Effect of Eliminating the 20-F Reconciliation on Analyst Following and Forecast Accuracy.

Dependent variable	Predicted sign	Log(Ana_follow)	Predicted sign	ACCURACY
Intercept		1.152**		2.372**
		(0.014)		(0.024)
TREAT (dummy)		−0.178		0.101
		(0.232)		(0.528)
TREAT×AFTER	−	−0.362***	−	0.111
		(<0.001)		(0.313)
AFTER (dummy)		0.360**		−0.617*
		(0.042)		(0.057)
LOGMV		0.016		−0.015
		(0.689)		(0.813)
BM		−0.350***		−0.710**
		(0.005)		(0.018)
Lev		−0.141		−1.787*
		(0.717)		(0.081)
STD_ROA		1.126		−0.520
		(0.332)		(0.714)
Sec_issue		0.063		0.266*
		(0.450)		(0.050)
Stock_ret		−0.038		0.346*
		(0.533)		(0.072)
UE		−0.836*		−6.800***
		(0.100)		(0.001)
CHROA		0.134		0.020
		(0.768)		(0.985)
EARNSKEW		0.133**		0.044
		(0.014)		(0.566)
RESTRUCT		0.363***		−0.242
		(0.000)		(0.145)
Log(Ana_follow)				0.226**
				(0.020)
Horizon				−1.557***
				(0.000)
Log(Firm_exp)				0.113**
				(0.044)
Log(Gen_exp)				0.011
				(0.858)
Log(Nfirm)				−0.091
				(0.311)
US				0.108
				(0.399)
Industry fixed effect	Yes		Yes
Year fixed effect	Yes		Yes
Adj. R² (%)	30.55		21.77
Number of obs.	492		1,726

Note. The regression with Log(Ana_follow) as the dependent variable is based on 492 firm-year observations. The regression with ACCURACY as the dependent variable is based on 1,726 firm-year-analyst observations. p values based on a two-tailed (one-tailed) t test are reported in parentheses for coefficients without predicted signs (for coefficients with predicted signs). Please refer to Supplemental Appendix for detailed variable definitions.

, **, and *** denote significance at the .10, .05, and .01 levels, respectively.

In the regression where the dependent variable is Log(Ana_follow), the coefficient of the interaction term, TREAT×AFTER, captures the difference in post-elimination change in analyst following between treatment firms and control firms. As predicted, it is negative (−0.362) and significant at the 1% level, using a one-tailed t test. Our results are consistent with the hypothesis that analyst coverage drops after the 20-F reconciliation becomes unavailable.¹² The coefficient on AFTER is positive and significant at the 5% level, suggesting that control firms are followed by more analysts in the post-elimination period.

When the dependent variable is ACCURACY, the coefficient on AFTER takes a negative sign, suggesting lower forecast accuracy for control firms in the post-elimination period. The coefficients on control variables are generally consistent with prior studies (Lang & Lundholm, 1996; Tan et al., 2011). The coefficient on Sec_issue and Stock_ret is positive and significant, suggesting that forecast accuracy is higher for firms with recent security issuance and high stock returns.¹³ Our interest is in the coefficient on the interaction term (TREAT×AFTER). It takes an opposite sign and is not significant at the 10% level, suggesting no change in forecast accuracy after the 20-F reconciliation is eliminated.¹⁴

In sum, the results in Table 2 show that eliminating the 20-F reconciliation significantly reduces analyst coverage but it seemingly has no impact on forecast accuracy.¹⁵

Testing H2

This section tests H2, which posits that the likelihood of terminating coverage of IFRS firms is higher for analysts who find the 20-F reconciliation useful than for other analysts.

We infer the usefulness of the 20-F reconciliation to analysts through the forecast revision around the 20-F filing date, when the 20-F reconciliation becomes publicly available. If financial analysts find the reconciliation useful, they are likely to revise their forecasts upon its public release.

Furthermore, the 20-F reconciliation provided by IFRS firms explains the differences between IFRS and U.S. GAAP. It is less useful to analysts who are familiar with both sets of rules. These analysts have a good understanding of the accounting rules and they may be able to estimate the accounting differences even if the 20-F reconciliation is unavailable. We infer analysts’ familiarity with the accounting rules through their coverage. Analysts who cover both IFRS firms and U.S. GAAP firms are therefore familiar with both sets of rules and the elimination of the 20-F reconciliation represents a less significant information loss to them. Consequently, we predict that these analysts are less likely to terminate their coverage of IFRS firms than other analysts, after the elimination of the 20-F reconciliation.

To test our predictions, we obtain all 159 firm-analyst observations from the treatment sample for the year prior to the elimination of the 20-F reconciliation. To determine whether analysts continue to provide coverage, we examine whether the same firm-analyst combination appears in the post-elimination period.

We use multivariate logistic regression for our empirical analysis. The dependent variable is STOP, a dummy which equals 1 if the analyst stops following the IFRS firm after the elimination of the 20-F reconciliation and 0 otherwise. The main independent variables are REVISION and FAM_BOTH. REVISION is a dummy variable indicating the occurrence of forecast revision around the 20-F filing date in the year prior to eliminating the 20-F reconciliation. It equals 1 if the forecast is revised in the 3-day window centered on the 20-F filing date, and 0 otherwise.¹⁶FAM_BOTH is a dummy variable indicating whether the analyst is familiar with both IFRS and U.S. GAAP. If an analyst covers both IFRS firms and U.S. GAAP firms, we assume that the analyst is familiar with both accounting rules and assign the value of 1 to FAM_BOTH. Otherwise, it has a value of 0.¹⁷ We additionally control for LOGMV, BM, Lev, STD_ROA, Sec_issue, Stock_ret, UE, CHROA, EARNSKEW, RESTRUCT, Horizon, Log(Firm_exp), Log(Gen_exp), Log(Nfirm), US and industry dummies. All independent variables are measured as of the year prior to eliminating the 20-F reconciliation. If H2 is true, we expect the coefficient on REVISION to be positive and significant and the coefficient on FAM_BOTH to be negative and significant.

Table 3 reports our regression results. In Regression 1, the coefficient on REVISION is 1.652, significant at the 5% level. The related odds ratio indicates that the odds of terminating coverage of the IFRS firm after the elimination of the 20-F reconciliation are more than four times higher for analysts who revise their forecasts around the 20-F filing date than for analysts who do not.

Table 3.

Analysts’ Decision to Terminate Coverage of IFRS Firms After the Elimination of the 20-F Reconciliation.

		Regression 1		Regression 2		Regression 3
Dependent variable: STOP	Predicted sign	Estimate	Odds ratio	Estimate	Odds ratio	Estimate	Odds ratio
Intercept		3.657		2.858		3.790
		(0.187)		(0.307)		(0.189)
REVISION (dummy)	+	1.652**	5.219			1.724**	5.607
		(0.039)				(0.036)
FAM_BOTH (dummy)	−			−2.185***	0.113	−2.245***	0.106
				(0.001)		(0.001)
LOGMV		−0.158	0.854	−0.167	0.846	−0.227	0.797
		(0.402)		(0.392)		(0.259)
BM		1.525	4.595	2.617	13.692	2.643	14.060
		(0.411)		(0.188)		(0.182)
Lev		6.999**	>999.999	7.929**	>999.999	7.276**	>999.999
		(0.012)		(0.011)		(0.018)
STD_ROA		−11.711 **	<0.001	−9.895**	<0.001	−11.175**	<0.001
		(0.012)		(0.045)		(0.028)
Sec_issue		−1.758*	0.172	−1.933*	0.145	−2.098**	0.123
		(0.064)		(0.061)		(0.044)
Stock_ret		−1.392	0.249	−1.472	0.230	−1.489	0.226
		(0.198)		(0.187)		(0.185)
UE		−4.796	0.008	−2.911	0.054	−2.760	0.063
		(0.536)		(0.719)		(0.727)
CHROA		8.597	>999.999	12.698	>999.999	11.890	>999.999
		(0.248)		(0.106)		(0.132)
EARNSKEW		−0.531*	0.588	−0.608**	0.544	−0.575*	0.563
		(0.061)		(0.038)		(0.051)
RESTRUCT		−0.505	0.604	−1.135*	0.322	−0.968	0.380
		(0.378)		(0.067)		(0.120)
Horizon		3.955*	52.187	4.312*	74.604	4.100*	60.322
		(0.077)		(0.059)		(0.079)
Log(Firm_exp)		−1.343 ***	0.261	−1.300***	0.273	−1.357***	0.258
		(0.000)		(0.001)		(<0.001)
Log(Gen_exp)		0.624**	1.865	0.687**	1.987	0.628*	1.874
		(0.047)		(0.036)		(0.057)
Log(Nfirm)		−0.874**	0.418	−0.087	0.917	−0.115	0.892
		(0.014)		(0.836)		(0.790)
US		−0.586	0.556	−0.376	0.686	−0.277	0.758
		(0.285)		(0.524)		(0.645)
Industry fixed effect		Yes				Yes
Max. Rescaled R² (%)		43.17		46.96		49.02
Number of obs.		159		159		159

Note. IFRS = International Financial Reporting Standards. This table reports results of logistic regressions. The dependent variable is STOP, which equals 1 if the analyst stops following the IFRS firm after the 20-f reconciliation becomes unavailable and 0 otherwise. p values based on a two-tailed (one-tailed) Wald chi-square test are reported in parentheses for coefficients without predicted signs (for coefficients with predicted signs). Please refer to Supplemental Appendix for detailed variable definitions.

, **, and *** denote significance at the .10, .05, and .01 levels, respectively.

In Regression 2, the coefficient on FAM_BOTH is −2.185, significant at the 1% level. The related odds ratio suggests that the odds of terminating coverage of the IFRS firm after eliminating the 20-F reconciliation are 88% lower for analysts who follow both IFRS firms and U.S. GAAP firms than other analysts.

In Regression 3, we include both REVISION and FAM_BOTH in one regression. The coefficient on REVISION is positive and significant at the 5% level, whereas the coefficient on FAM_BOTH is negative and significant at the 1% level.

Overall, Table 3 shows that the likelihood of terminating coverage of IFRS firms is higher for analysts who revise their forecasts around the 20-F filing date and lower for analysts who cover both IFRS firms and U.S. GAAP firms. This finding is consistent with the notion that analysts who are affected more by the elimination of the 20-F reconciliation are more likely to terminate their coverage of IFRS firms.

Testing H3

H3 posits that eliminating the 20-F reconciliation results in a more pronounced drop in analyst following and forecast accuracy for firms whose 20-F reconciliation is more informative.

We use four measures of the informativeness of the reconciliation. The first measure is the likelihood that analysts revise their forecasts around the 20-F filing date. A higher likelihood shows that analysts perceive the reconciliation to be more useful. The second measure is the magnitude of the abnormal return around the 20-F filing date. A greater magnitude shows that investors react more dramatically to the 20-F reconciliation, implying that the 20-F reconciliation is more informative. The third measure is the absolute difference between IFRS earnings and U.S. GAAP earnings. A wider difference renders it more difficult to infer one earnings number from the other, implying that the reconciliation is more informative. The last measure is the number of reconciliation items between the IFRS earnings and the GAAP earnings. Conceivably, a greater number of items reflects a more convoluted reconciliation process and the reconciliation is thus more useful.¹⁸

Empirically, we use H_rev to measure the likelihood of analysts revising their forecasts. H_rev is a dummy variable, which equals 1 if the average percentage of analysts who revise their forecasts in the 3-day window centered on the 20-F filing date in the pre-elimination period is above the treatment sample median, and 0 otherwise.

We use H_CAR to measure the magnitude of investors’ response to the 20-F reconciliation. H_CAR is a dummy variable, which equals 1 if the average magnitude of the 3-day cumulative abnormal return (CAR) centered on the 20-F filing date in the pre-elimination period is above the treatment sample median, and 0 otherwise.¹⁹

H_diff measures the divergence between the GAAP and IFRS earnings. It equals 1 if the average absolute reconciliation difference (defined as the absolute value of the difference between U.S. GAAP earnings and IFRS earnings deflated by IFRS earnings) in the pre-elimination period is above the treatment sample median, and 0 otherwise.

H_items is an indicator for the number of reconciliation items. It equals 1 if the average number of reconciliation items in the pre-elimination period is above the treatment sample median, and 0 otherwise.

To test our hypothesis, we first divide the treatment firms into two subsamples according to the value of H_rev/H_CAR/H_diff/H_items. We then rerun Models (1) and (2), using all control firms and one of the subsamples. The coefficient on TREAT×AFTER indicates the difference in the post-elimination change between the particular subsample and control firms. As we have two subsamples, we obtain two sets of coefficient estimates. We directly compare the coefficients on TREAT×AFTER between two subsamples.

Our regression results are reported in Table 4. Panel A reports the regression results when subsamples are formed according to H_rev. In the analyst following regression, we find that the coefficient on TREAT×AFTER is −0.402, significant at the 1% level, when treatment firms include only observations with H_rev being equal to 1 (i.e., firms whose analysts are more likely to revise their forecasts around the 20-F filing date); it is however insignificant at the 10% level, when treatment firms include only observations with being H_rev equal to 0. This finding reveals that the significant reduction in analyst following mainly comes from treatment firms whose analysts are more likely to revise their forecasts around the 20-F filing date. Our results indicate that the drop in analyst coverage is significantly more pronounced for firms whose analysts are more likely to revise their forecasts around the 20-F filing date.

Table 4.

The Effect of Eliminating the 20-F Reconciliation: Subsample Analysis.

Panel A: High and Low Revision Likelihood.
	Log(Ana_follow)			ACCURACY
Dependent variable	Predicted sign	H_rev = 1 subsample	H_rev = 0 subsample	Predicted sign	H_rev = 1 subsample	H_rev = 0 subsample
Intercept		1.059	0.991*		0.770	2.159*
		(0.102)	(0.051)		(0.444)	(0.050)
TREAT (dummy)		−0.252	−0.271*		−0.059	0.135
		(0.253)	(0.075)		(0.767)	(0.490)
TREAT×AFTER	−	−0.402***	−0.170	−	−0.060*	0.312
		(0.001)	(0.200)		(0.085)	(0.222)
AFTER (dummy)		0.377**	0.370**		−0.622*	−0.601*
		(0.041)	(0.028)		(0.073)	(0.074)
LOGMV		0.017	0.011		0.008	−0.001
		(0.742)	(0.785)		(0.905)	(0.989)
BM		−0.314**	−0.270**		−0.629**	−0.646**
		(0.020)	(0.028)		(0.047)	(0.038)
Lev		−0.159	0.024		−1.833*	−1.876*
		(0.719)	(0.950)		(0.090)	(0.092)
STD_ROA		0.849	1.776		−0.158	−0.339
		(0.551)	(0.198)		(0.922)	(0.851)
Sec_issue		0.026	0.115		0.221	0.238*
		(0.799)	(0.184)		(0.128)	(0.099)
Stock_ret		−0.068	−0.014		0.383*	0.287
		(0.318)	(0.831)		(0.075)	(0.164)
UE		−0.692	−1.199**		−8.276***	−7.017***
		(0.303)	(0.016)		(0.001)	(0.004)
CHROA		0.390	0.104		0.131	0.981
		(0.397)	(0.812)		(0.908)	(0.484)
EARNSKEW		0.162**	0.108**		0.054	0.055
		(0.023)	(0.045)		(0.531)	(0.504)
RESTRUCT		0.398***	0.247**		−0.167	−0.263
		(0.001)	(0.011)		(0.374)	(0.198)
Log(Ana_follow)					0.194*	0.201*
					(0.054)	(0.071)
Horizon					−1.649***	−1.513***
					(0.000)	(0.001)
Log(Firm_exp)					0.113*	0.133*
					(0.073)	(0.052)
Log(Gen_exp)					0.017	−0.018
					(0.787)	(0.800)
Log(Nfirm)					−0.091	−0.095
					(0.365)	(0.295)
US					0.127	0.102
					(0.387)	(0.418)
Industry fixed effect		Yes	Yes		Yes	Yes
Year fixed effect		Yes	Yes		Yes	Yes
Adj. R² (%)		31.11	27.35		23.27	22.00
Number of obs.		404	420		1,500	1,437
Panel B: High and Low Investors’ Reaction to the 20-F Reconciliation.
Dependent variable	Log(Ana_follow)			ACCURACY
	Predicted sign	H_CAR = 1 subsample	H_CAR = 0 subsample	Predicted sign	H_CAR = 1 subsample	H_CAR = 0 subsample
Intercept		1.061**	0.173		2.523**	0.184
		(0.047)	(0.746)		(0.024)	(0.833)
TREAT (dummy)		−0.151	−0.364*		0.101	0.086
		(0.424)	(0.059)		(0.589)	(0.689)
TREAT×AFTER	−	−0.338***	−0.157	−	−0.034*	0.283
		(0.008)	(0.246)		(0.090)	(0.159)
AFTER (dummy)		0.366**	0.368**		−0.562*	−0.665*
		(0.042)	(0.032)		(0.090)	(0.054)
LOGMV		0.010	−0.002		0.005	−0.008
		(0.828)	(0.968)		(0.943)	(0.901)
BM		−0.257*	−0.343***		−0.586*	−0.686**
		(0.055)	(0.006)		(0.065)	(0.027)
Lev		−0.020	−0.086		−1.969*	−1.720
		(0.962)	(0.836)		(0.064)	(0.135)
STD_ROA		1.635	0.858		−0.410	−1.208
		(0.272)	(0.509)		(0.805)	(0.503)
Sec_issue		0.056	0.081		0.266*	0.213
		(0.521)	(0.405)		(0.057)	(0.156)
Stock_ret		−0.012	−0.057		0.363*	0.280
		(0.866)	(0.372)		(0.074)	(0.190)
UE		−0.897	−0.772		−8.676***	−6.636***
		(0.165)	(0.132)		(0.001)	(0.005)
CHROA		0.488	0.102		0.372	0.636
		(0.297)	(0.804)		(0.799)	(0.558)
EARNSKEW		0.115*	0.147**		0.060	0.031
		(0.073)	(0.019)		(0.470)	(0.727)
RESTRUCT		0.285***	0.374***		−0.306	−0.199
		(0.006)	(0.000)		(0.113)	(0.314)
Log(Ana_follow)					0.229**	0.188*
					(0.035)	(0.062)
Horizon					−1.661***	−1.511
					(0.001)	(0.000)
Log(Firm_exp)					0.122*	0.110*
					(0.070)	(0.085)
Log(Gen_exp)					0.020	−0.027
					(0.769)	(0.673)
Log(Nfirm)					−0.164*	−0.022
					(0.091)	(0.812)
US					0.205	0.037
					(0.160)	(0.770)
Industry fixed effect		Yes	Yes		Yes	Yes
Year fixed effect		Yes	Yes		Yes	Yes
Adj. R² (%)		27.34	29.67		24.10	21.23
Number of obs.		416	408		1,449	1,488
Panel C: High and Low Earnings Difference.
Dependent variable	Log(Ana_follow)			ACCURACY
	Predicted sign	H_diff = 1 subsample	H_diff = 0 subsample	Predicted sign	H_diff = 1 subsample	H_diff = 0 subsample
Intercept		1.450***	0.951*		0.543	2.290**
		(0.002)	(0.072)		(0.516)	(0.030)
TREAT (dummy)		−0.382**	−0.136		0.030	0.068
		(0.033)	(0.442)		(0.880)	(0.775)
TREAT×AFTER	−	−0.427***	−0.157	−	−0.134*	0.236
		(0.002)	(0.208)		(0.073)	(0.160)
AFTER (dummy)		0.402**	0.342**		−0.628*	−0.607*
		(0.026)	(0.044)		(0.067)	(0.068)
LOGMV		0.010	0.016		0.000	−0.012
		(0.826)	(0.729)		(0.997)	(0.854)
BM		−0.262**	−0.326**		−0.604*	−0.665**
		(0.035)	(0.012)		(0.055)	(0.031)
Lev		0.015	0.032		−1.897*	−1.717
		(0.971)	(0.939)		(0.093)	(0.111)
STD_ROA		1.921	0.798		−0.804	−0.339
		(0.194)	(0.549)		(0.683)	(0.828)
Sec_issue		0.079	0.049		0.239	0.232
		(0.405)	(0.589)		(0.104)	(0.101)
Stock_ret		−0.015	−0.074		0.311	0.318
		(0.817)	(0.290)		(0.129)	(0.128)
UE		−0.898	−0.568		−6.503***	−8.926***
		(0.102)	(0.358)		(0.004)	(0.002)
CHROA		0.497	0.020		0.338	0.848
		(0.289)	(0.963)		(0.810)	(0.446)
EARNSKEW		0.115*	0.131**		0.020	0.071
		(0.058)	(0.037)		(0.812)	(0.400)
RESTRUCT		0.350***	0.328***		−0.251	−0.238
		(0.002)	(0.002)		(0.245)	(0.165)
Log(Ana_follow)					0.200*	0.198**
					(0.060)	(0.049)
Horizon					−1.602***	−1.510***
					(0.002)	(0.000)
Log(Firm_exp)					0.141*	0.096
					(0.056)	(0.111)
Log(Gen_exp)					−0.012	−0.002
					(0.863)	(0.974)
Log(Nfirm)					−0.176	−0.026
					(0.100)	(0.765)
US					0.181	0.069
					(0.260)	(0.562)
Industry fixed effect		Yes	Yes		Yes	Yes
Year fixed effect		Yes	Yes		Yes	Yes
Adj. R² (%)		29.73	28.36		22.41	23.09
Number of obs.		408	416		1,387	1,550
Panel D: High and Low Numbers of Reconciliation Items.
Dependent variable	Log(Ana_follow)			ACCURACY
	Predicted sign	H_items = 1 subsample	H_items = 0 subsample	Predicted sign	H_items = 1 subsample	H_items = 0 subsample
Intercept		1.356***	1.070*		0.599	2.160**
		(0.003)	(0.053)		(0.460)	(0.048)
TREAT (dummy)		−0.120	−0.369**		−0.028	0.082
		(0.552)	(0.041)		(0.863)	(0.726)
TREAT×AFTER	−	−0.411***	−0.077	−	−0.046**	0.304
		(0.001)	(0.288)		(0.041)	(0.105)
AFTER (dummy)		0.375**	0.370**		−0.641*	−0.610*
		(0.033)	(0.035)		(0.064)	(0.070)
LOGMV		−0.001	0.022		−0.008	−0.003
		(0.975)	(0.642)		(0.906)	(0.962)
BM		−0.267**	−0.317**		−0.558*	−0.706**
		(0.041)	(0.013)		(0.080)	(0.022)
Lev		0.015	−0.123		−1.880*	−1.749
		(0.971)	(0.766)		(0.084)	(0.125)
STD_ROA		0.964	1.672		−1.253	−0.190
		(0.493)	(0.233)		(0.479)	(0.911)
Sec_issue		0.065	0.075		0.220	0.257*
		(0.489)	(0.421)		(0.111)	(0.087)
Stock_ret		−0.024	−0.037		0.340*	0.286
		(0.745)	(0.549)		(0.099)	(0.167)
UE		−0.495	−1.137**		−7.931***	−7.159***
		(0.424)	(0.031)		(0.002)	(0.004)
CHROA		0.398	0.153		0.533	0.656
		(0.408)	(0.702)		(0.699)	(0.562)
EARNSKEW		0.138**	0.118**		0.014	0.063
		(0.040)	(0.049)		(0.872)	(0.447)
RESTRUCT		0.274**	0.359***		−0.218	−0.256
		(0.016)	(0.002)		(0.273)	(0.185)
Log(Ana_follow)					0.190*	0.205*
					(0.058)	(0.056)
Horizon					−1.538***	−1.547***
					(0.001)	(0.000)
Log(Firm_exp)					0.120*	0.117*
					(0.074)	(0.072)
Log(Gen_exp)					0.001	−0.024
					(0.988)	(0.725)
Log(Nfirm)					−0.129	−0.066
					(0.202)	(0.465)
US					0.128	0.118
					(0.386)	(0.355)
Industry fixed effect		Yes	Yes		Yes	Yes
Year fixed effect		Yes	Yes		Yes	Yes
Adj. R² (%)		28.96	28.22		22.64	22.64
Number of obs.		416	408		1,474	1,463

Note. IFRS = International Financial Reporting Standards; GAAP = Generally Accepted Accounting Principles. This table reports results from testing the effect of eliminating IFRS–U.S. GAAP reconciliations on analyst following and analyst forecast accuracy in subsamples of treatment firms. p values based on a two-tailed (one-tailed) t test are reported in parentheses for coefficients without predicted signs (for coefficients with predicted signs).

, **, and *** denotes significance at the .10, .05, and .01 levels, respectively.

In the forecast accuracy regression, we find that the coefficient on TREAT×AFTER is −0.060 and significant at the 10% level when treatment firms include only observations with H_rev being equal to 1. When treatment firms include only observations with H_rev being equal to 0, the coefficient on TREAT×AFTER is 0.312 and insignificant. This finding indicates that, contrary to the conclusion based on the overall sample, forecast accuracy drops significantly for firms whose analysts are more likely to revise their forecasts around the 20-F filing date.

Panels B to D report the regression results when subsamples are formed according to H_CAR/H_diff/H_items. Results are qualitatively similar to those reported in Panel A. We find that the significant reduction in analyst following mainly comes from treatment firms with H_CAR/H_diff/H_items being equal to 1 (i.e., firms whose investors react more strongly to the 20-F reconciliation, firms whose GAAP earnings diverge to a greater extent from IFRS earnings, and firms whose reconciliations include a greater number of items). Different from the conclusion based on the overall sample, the drop in forecast accuracy is more pronounced for firms with more informative 20-F reconciliations.

In sum, Table 4 shows that eliminating the 20-F reconciliation results in a more pronounced drop in analyst following and forecast accuracy for firms with more informative 20-F reconciliations.²⁰ Contrary to the insignificant finding based on the overall sample, for these firms, forecast accuracy deteriorates significantly after the elimination. These results highlight the importance of analyzing subsamples of firms for a complete assessment of the impact of eliminating the 20-F reconciliation.

Conclusion

In November 2007, the SEC allowed FPIs to file their financial statements without reconciling to the U.S. GAAP, if they follow IFRS as adopted by IASB. This decision has been controversial both in academia and among professional organizations.

This study investigates how the elimination of the 20-F reconciliation affects financial analysts. Financial analysts are regarded as sophisticated and influential market participants and the literature suggests that their actions have valuation consequences. Therefore, we analyze the impact of the SEC’s rule from an important and relevant perspective.

We find that, in general, eliminating the 20-F reconciliation significantly reduces analyst following but it has no impact on forecast accuracy. We conjecture that this seemingly inconsistent finding is due to analysts’ choice: analysts who are affected by the elimination are likely to terminate their coverage, whereas those who are not affected by the elimination are likely to continue their coverage.

We find that the odds of terminating coverage of the IFRS firm after the elimination of the 20-F reconciliation are more than four times higher for analysts who revise their forecasts around the 20-F filing date than those for analysts who do not, whereas the odds are 88% lower for analysts who cover both IFRS firms and U.S. GAAP firms than those for analysts who do not. Our results suggest that analysts who are affected by the elimination of the 20-F reconciliation to a greater extent are more likely to drop their coverage and become empirically unobservable. Our findings are consistent with the notion that analyses based on observable forecast data are biased toward concluding that eliminating the 20-F reconciliation has no impact on financial analysts.

We next hypothesize that eliminating the 20-F reconciliation results in a more pronounced drop in analyst following and forecast accuracy for firms with more informative 20-F reconciliations. We use four measures of the informativeness of the reconciliation and results from all four measures support our hypothesis. In contrast to the insignificant finding based on the overall sample, we show that forecast accuracy deteriorates significantly for firms whose 20-F reconciliations are useful. These results highlight the importance of analyzing subsamples of firms for a comprehensive assessment of the impact of eliminating the 20-F reconciliation.

Overall, our results suggest that eliminating the 20-F reconciliation imposes costs on financial analysts. Our study also provides policy implications for SEC’s recently proposed rules to simplify disclosure requirements for public companies.

Supplemental Material

Online_Appendix – Supplemental material for Analyst Following and Forecast Accuracy After Eliminating the 20-F Reconciliation Between IFRS and U.S. GAAP

Supplemental material, Online_Appendix for Analyst Following and Forecast Accuracy After Eliminating the 20-F Reconciliation Between IFRS and U.S. GAAP by Wen Li and Huai Zhang in Journal of Accounting, Auditing & Finance

Footnotes

Acknowledgements

The authors acknowledge helpful comments from Alastair Lawrence, Clive Lennox, Bin Ke, and seminar participants at Nanyang Technological University, Shanghai Jiao Tong University, and the 2013 AAA Financial Accounting and Reporting Section (FARS) Midyear Meeting. Their sincere gratitude goes to Hongping Tan, who kindly shared his analyst location data with them.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: H.Z. acknowledges the financial support from Singapore Ministry of Education Academic Research Fund Tier 1 (Reference No. RG163/17).

ORCID iDs

Wen Li

Huai Zhang

Supplemental Material

Supplemental material for this article is available online.

Notes

References

Ashbaugh

Pincus

(2001). Domestic accounting standards, international accounting standards, and the predictability of earnings. Journal of Accounting Research, 39, 417–434.

Bae

Tan

Welker

(2008). International GAAP differences: The impact on foreign analysts. The Accounting Review, 83, 593–628.

Barth

Kasznik

Mcnichols

(2001). Analyst coverage and intangible assets. Journal of Accounting Research, 39, 1–34.

Bertrand

Duflo

Mullainathan

(2004). How much should we trust differences-in-differences estimates? Quarterly Journal of Economics, 119, 249–275.

Beyer

Cohen

Lys

T. Z.

Walther

B. R.

(2010). The financial reporting environment: Review of the recent literature. Journal of Accounting and Economics, 50, 296–343.

Bhushan

(1989). Firm characteristics and analyst following. Journal of Accounting and Economics, 11, 255–274.

Brown

(2001). Predicting individual analyst earnings forecast accuracy. Financial Analysts Journal, 57, 44–49.

Byard

(2011). The effect of mandatory IFRS adoption on financial analysts’ information environment. Journal of Accounting Research, 49, 69–96.

Byard

Mashruwala

Suh

(2010). The economic consequences of eliminating the reconciliation of IFRS to U.S.GAAP: An information transfer analysis [Working paper]. City University of New York.

10.

CFA Institute. (2007). Comment letter to SEC dated October 2, 2007 (Re: File No. S7-13-07). http://sec.gov/comments/s7-13-07/s71307-125.pdf

11.

Chaney

P. K.

Hogan

C. E.

Jeter

D. C.

(1999). The effect of reporting restructuring charges on analysts’ forecast revisions and errors. Journal of Accounting and Economics, 27, 261–284.

12.

Chen

Deng

Gupta

Sami

(2015). The SEC’s elimination of 20-F reconciliation and information asymmetry. Journal of Contemporary Accounting and Economics, 11, 75–87.

13.

Chen

Khurana

I. K.

(2015). The impact of eliminating the Form 20-F reconciliation on shareholder wealth: Evidence from U.S. cross-listed firms. The Accounting Review, 90(1), 199–228.

14.

Chen

Krishnan

Sami

(2015). Goodwill impairment charges and analyst forecast properties. Accounting Horizons, 29, 141–169.

15.

Chen

Sami

(2008). Trading volume reaction to the earnings reconciliation from IAS to U.S. GAAP. Contemporary Accounting Research, 25, 15–33.

16.

Chen

Sami

(2010). Trading volume reaction to the earnings reconciliation from IFRS to U.S.GAAP: Further evidence [Working paper]. Arizona State University; Lehigh University.

17.

Clement

(1999). Analyst forecast accuracy: Do ability, resources, and portfolio complexity matter? Journal of Accounting and Economics, 27, 285–303.

18.

Clement

Tse

(2005). Financial analyst characteristics and herding behavior in forecasting. The Journal of Finance, 60, 307–341.

19.

De Franco

D. G.

Kothari

S. P.

Verdi

R. S

. (2011). The benefits of financial statement comparability. Journal of Accounting Research, 49, 895–931.

20.

Duru

Reeb

D. M.

(2002). International diversification and analysts’ forecast accuracy and bias. The Accounting Review, 77, 415–433.

21.

Gordon

Jorgensen

Linthicum

(2009). Could IFRS replace U.S.GAAP? A comparison of earnings attributes and the informativeness in the U.S. market [Working paper]. Temple University; Columbia University; University of Texas at San Antonio.

22.

Guan

Hope

O.-K.

Kang

(2006). Does similarity of local GAAP to U.S. GAAP explain analysts’ forecast accuracy? Journal of Contemporary Accounting and Economics, 2(2), 151–169.

23.

Hansen

Pownall

Prakash

Vulcheva

(2010). Relaxing the reconciliation requirement in non-U.S. firms’ SEC filings: Changes in GAAP, earnings attributes, and earnings informativeness [Working paper]. Emory University.

24.

Harris

Muller

(1999). The market valuation of IAS versus U.S. GAAP accounting measures using form 20-F reconciliations. Journal of Accounting and Economics, 26, 285–312.

25.

Henry

Lin

Yang

(2009). The European-U.S. GAAP gap: Amount, type, homogeneity and value relevance of IFRS to U.S. GAAP from 20-F reconciliations. Accounting Horizons, 23, 121–150.

26.

Jacob

Lys

Neale

(1999). Expertise in forecasting performance of security analysts. Journal of Accounting and Economics, 28, 27–50.

27.

Jiang

Petroni

K. R.

Wang

I. Y.

(2010). Did eliminating the 20-F reconciliation between IFRS and US GAAP matter? [Working paper]. Michigan State University.

28.

Kang

Krishnan

G. V.

Wolfe

M. C.

H. S.

(2012). The impact of eliminating the 20-F reconciliation requirement for IFRS filers on earnings persistence and information uncertainty. Accounting Horizons, 26(4), 741–765.

29.

Kim

(2012). Does eliminating the Form 20-F reconciliation from IFRS to U.S. GAAP have capital market consequences? Journal of Accounting and Economics, 53, 249–270.

30.

Lang

M. H.

Lins

K. V.

Miller

D. P.

(2004). Concentrated control, analyst following, and valuation: Do analysts matter most when investors are protected least? Journal of Accounting Research, 42, 589–623.

31.

Lang

M. H.

Lundholm

(1996). Corporate disclosure policy and analyst behavior. The Accounting Review, 71, 467–492.

32.

Lin

Y.-H.

Huang

H.-W.

(2014). Eliminating the Form 20-F reconciliation and audit pricing [Working paper]. National Cheng Kung University.

33.

Mikhail

Walther

Willis

(1997). Do security analysts improve their performance with experience? Journal of Accounting Research, 35, 131–166.

34.

O’Brien

P. C.

Bhushan

(1990). Analyst following and institutional ownership. Journal of Accounting Research, 28, 55–76.

35.

Plumlee

(2007). 20-F filers and SEC proposed changes: Some evidence of a U.S. Home GAAP preference [Working paper]. University of Utah.

36.

Rock

Sedo

Willenborg

(2001). Analyst following and count-data econometrics. Journal of Accounting and Economics, 30, 351–373.

37.

Stickel

(1991). Common stock returns surrounding earnings forecast revisions: More puzzling evidence. The Accounting Review, 66, 402–416.

38.

Tan

Wang

Welker

(2011). Analyst following and forecast accuracy after mandated IFRS adoptions. Journal of Accounting Research, 49, 1307–1356.

39.

U.S. Securities and Exchange Commission. (2005). A securities regulator looks at convergence (Ed., Nicolaisen

Donald T.

). http://www.sec.gov/news/speech/spch040605dtn.htm

40.

Yang

Henry

(2013). Does multistandard reporting affect analysts’ forecasts? Journal of Corporate Accounting & Finance, 24, 55–67.

41.

(2011). The Interaction of voluntary and mandatory disclosures: Evidence from the SEC’s elimination of the IFRS-U.S.GAAP reconciliation [Working paper]. University of Georgia.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.34 MB